Volume 14 Issue 3

A Multi-Branch Deepfake Detection Framework Using Spatial, Noise, And Frequency Domain Features With Attention Fusion

Authors: Lakshay Bhardwaj, Rishabh Jain, Kritika, Ritesh Kumar

Abstract: Deepfakes have emerged as a major challenge in digital media forensics because modern generative models can produce highly realistic fake facial content that is difficult to distinguish from authentic media. Their rapid growth has increased the risk of misinformation, identity misuse, and security threats in online environments, motivating the need for reliable forensic datasets and detection frameworks [1], [7], [10]. Existing deepfake detection methods often perform well only under controlled settings and show limited robustness when evaluated on unseen manipulations or datasets. Many CNN-based methods achieve high accuracy on known datasets but fail to generalize to unseen data. Prior works based on compact CNNs, frequency-aware learning, and multi-branch detection have shown promising performance, but cross-dataset generalization remains a major challenge [2]–[5], [8]. To address this issue, this work proposes a multi-branch deepfake detection framework that jointly learns from spatial appearance information, residual noise traces, and frequency-domain artifacts. The spatial branch uses a pretrained EfficientNet-B0 backbone to capture facial inconsistencies [6], the noise branch extracts forensic residual cues using SRM-based filtering inspired by image manipulation detection methods [9], and the frequency branch analyzes the log magnitude spectrum obtained through FFT transformation to reveal spectral anomalies commonly associated with forged content [3]. An attention-based fusion module combines these complementary representations and adaptively emphasizes the most discriminative branch for each sample, following the motivation of prior multi-domain and multi-branch approaches [4], [5]. The model is trained and evaluated on the FaceForensics++ dataset using frame-level samples derived from video sequences [1]. Experimental results show that the proposed framework achieves a final test accuracy of 63.75%, demonstrating that multi-domain feature fusion is effective for improving deepfake detection performance. The results further indicate that attention-guided fusion helps the classifier exploit complementary forensic evidence beyond conventional RGB-only models.

DOI: http://doi.org/10.5281/zenodo.20199710

Ethnomedicinal Plants And Indigenous Knowledge System Of A Rural Community In Iligan City, Philippines

Authors: Samson L. Mangin, Edna B. Nabua

Abstract: Ethnomedicinal knowledge remains an important component of primary healthcare in many rural communities, yet it is increasingly threatened by modernization and environmental change. This study documents the ethnomedicinal plants used by the community of Sitio Langilanon, Barangay Pugaan, Iligan City, and examines their cultural and therapeutic significance. A mixed-methods research design was employed, integrating structured surveys and semi-structured interviews among selected settlers of Sitio Langilanon. Data collection focused on identifying medicinal plant species, ailments treated, methods of preparation, and associated indigenous health practices. Qualitative data were thematically analyzed, while quantitative data were summarized using descriptive statistics. Twenty-two (22) ethnomedicinal plant species were documented. Common methods of preparation included boiling, pounding, and infusion. These remedies were primarily used to treat ailments such as fever, diarrhea, cough, wounds, and hypertension. Findings revealed a strong community reliance on plant-based medicine and demonstrated the continued intergenerational transmission of indigenous medicinal knowledge. The study highlights the vital role of ethnomedicine in the local healthcare system of Sitio Langilanon. However, traditional medicinal practices face growing threats from cultural shifts and environmental degradation. Systematic documentation, cultural preservation initiatives, and scientific validation are essential to safeguard this knowledge and enhance its contribution to sustainable healthcare and broader scientific research.

DOI: http://doi.org/10.5281/zenodo.20200858

Biomertic Based MultiFactor Authentatication Using Behavioural Analysis

Authors: Ramineni Teja, Sankalp Bhawsar

Abstract: Passwords alone are no longer enough to keep systems secure in today’s rapidly changing cybersecurity land- scape. This study explores a smarter approach using biometric based multifactor authentication (MFA), focusing on how people interact with devices such as their typing patterns, mouse move- ments, and touch behavior. By combining these behavioral traits with machine learning, the system can continuously verify users with over 97 Percentage accuracy, without interrupting their experience. The research also looks at important challenges like preventing spoofing attacks, protecting user privacy, and ensuring the system can scale effectively. Overall, the findings suggest that behavioral biometrics can play a key role in building more secure and userfriendly authentication systems for the future.

DOI: http://doi.org/10.5281/zenodo.20201223

Smart Nutribot:A Web Application For Personalized Dietary Recommender Using Xgboost And Random Forest.

Authors: Jettty Varshitha, Gottipati Abhinaya, Kolusu Ankitha, Ms. S.A. Neelavani

Abstract: Smart Nutribot is an intelligent web- based application designed to provide personalized dietary recommendations using advanced machine learning techniques such as XGBoost and Random Forest. The system aims to address the growing need for customized nutrition plans by analyzing individual user data, including age, gender, weight, health conditions, dietary preferences, and lifestyle habits. By leveraging the predictive capabilities of ensemble learning models, the application generates accurate and tailored meal suggestions that promote healthy living and disease prevention. The web interface ensures user- friendly interaction, allowing users to input their details and receive instant recommendations in an accessible format. XGBoost enhances the model’s performance through efficient gradient boosting, while Random Forest improves robustness by reducing overfitting and increasing prediction accuracy. The integration of these models enables Smart Nutribot to deliver reliable and data-driven dietary guidance. Overall, this project demonstrates the practical application of machine learning in healthcare and nutrition, providing a scalable and efficient solution for personalized diet planning.

Comparative Study of Oxidative Stress Response Between Gram-positive and Gram-negative Bacteria

Authors: Sittie Fatmah M. Abdulrahman, Abdulraffy A. Alikhan, Najibah A. Casim, Janisah O. Hadji Amen, Samson L. Mangin, Junge B. Guillena, Doyne Grace Laparan, Sweet Maraesol Cabrera

Abstract: This study compared the oxidative stress responses of Gram-positive and Gram-negative bacteria exposed to hydrogen peroxide (5–20 mM) and a vitamin C–ferric chloride (FeCl₃) system under varying environmental conditions. Using the disc diffusion method, zones of inhibition were measured for Staphylococcus aureus, Streptococcus pyogenes, Pseudomonas aeruginosa, and Salmonella spp. Gram-negative bacteria were generally more susceptible to hydrogen peroxide, particularly under aerobic conditions, while Gram-positive bacteria showed greater resistance. The vitamin C–FeCl₃ system demonstrated consistent broad-spectrum antibacterial activity across all strains. Environmental factors, especially oxygen availability, influenced oxidative susceptibility. Findings highlight structural and physiological differences affecting bacterial oxidative stress tolerance and suggest the potential of vitamin C–FeCl₃ as an adjunct antimicrobial strategy.

DOI: https://doi.org/10.5281/zenodo.20201650

AI Based Crop Prediction And Recommendation

Authors: Sanskar Kadam, Varad Jamdar, Krushna Kapse, Snehal Shere, Shrawani Mule

Abstract: Farming is a big part of India's economy, making up about 17% of the country's GDP. It also helps support the lives of more than half of the people living in rural areas. But even with its importance, many Indian farmers face crop failures every year. This often happens because they don't have access to affordable tools that can give them good advice on which crops to plant, based on scientific research. To solve this problem, we've developed a system that uses artificial intelligence to predict and recommend crops. It combines special hardware that senses the condition of the soil with a type of machine learning model that considers many factors at once. This allows the system to give farmers personalized advice in real time, helping them make informed decisions about which crops to plant. Here's a rewritten version of the input text in a more human-like tone, similar to the provided reference human samples: When it comes to measuring soil nutrients, we've developed a hardware prototype that's pretty impressive. It's made up of an RS-485 Modbus NPK sensor, a DHT11 temperature and humidity sensor, a capacitive soil moisture sensor, and an Arduino UNO microcontroller. In the early stages, we even experimented with a TDS sensor as a low-cost alternative for estimating soil nutrient characteristics. This approach helped us keep costs down while still testing the system's architecture. But in the end, we decided to use actual NPK sensor readings as the primary input for soil nutrients. We also trained and compared seven different machine learning models using a dataset of 2,200 agricultural samples covering 22 different crop classes. These models included Random Forest, XGBoost, LightGBM, SVM, Gradient Boosting, KNN, and Logistic Regression. And what we found was that a soft-voting ensemble combining Random Forest, XGBoost, and LightGBM achieved an impressive 99.77% test accuracy and 99.73% mean cross-validation accuracy. But here's the thing: we didn't just stop at soil nutrients. We also incorporated real-time weather data for temperature, humidity, and rainfall into our model, using the OpenWeatherMap API. This allows us to provide location-aware recommendations that take into account the specific weather conditions in a given area. And the best part? The entire system is deployed as a user-friendly Gradio web application, with three different output tabs: crop recommendation with confidence bars, soil health analysis with fertiliser advice, and a seasonal crop planner. What's really exciting about this system is that it directly supports the United Nations' Sustainable Development Goal 2 – Zero Hunger. By providing farmers with accurate and reliable recommendations, we can help increase crop yields and reduce hunger around the world. It's a big goal, but we're hopeful that our system can make a real difference.

DOI: https://doi.org/10.5281/zenodo.20204819

Self-Attention–Driven Vision Transformer Model For Autism Identification And Cognitive Skill Enhancement

Authors: M. Menakapriya, M. Rasika, G. Prabha, R. Lavanya, S. Mounika, V. Nagarajan

Abstract: Early identification and treatment of Autism Spectrum Disorder (ASD) pose difficulties due to its complex neurological nature and heterogeneous symptoms. In this work, an innovative Self-Attention-Driven Vision Transformer (SA-ViT) model is introduced to cater to both ASD detection and cognitive skills improvement within one approach. Our work benefits from self-attention properties of vision transformers to detect subtle patterns in the behavior of ASD individuals as well as generate structured cognitive stimulation material. By extracting features from facial imagery, videos, and behavioral data, our SA-ViT can classify ASD samples with 97.6% accuracy on the ASD Facial Image Dataset, which beats regular CNN models (91.2%) and regular ViTs (94.8%). In terms of cognitive skills improvement, we were able to develop personalized structured tasks that resulted in 34.2% improved visual memory retention and 28.7% enhanced pattern recognition after eight weeks. The use of explainable AI approaches (Grad-CAM) enhances our system's applicability in a medical setting.

DOI: https://doi.org/10.5281/zenodo.20205889

Real-World Case Study: Evaluating AI-Powered And Traditional Signature Approaches To Email Phishing Threats

Authors: Shrushti Kaza, Akhila Harshini Gadamsetty, Abhijeet Raj, Pranav Veer Singh, Dr M Umamaheswari

Abstract: Email-based phishing is among the persistent and costly cybersecurity challenges which exploit human gullibility, social engineering, and organizational frailties for gaining un-approved access to sensitive information. Signature-based cyber-security strategies use preset patterns, blacklists, and heuristic approaches to identify phishing emails. While signature-based detection systems can recognize phishing emails with known char-acteristics successfully, they usually fail to identify sophisticated attacks which evade recognition due to their novelty or disguise. On the other hand, modern technologies based on ML and NLP employ numerous features including email body, natural language used, sender behavior, URLs embedded in an email, and additional metadata. The ability of such approaches to generalize makes them applicable in the detection of previously unseen phishing campaigns. In this study, comprehensive comparison between AI-powered phishing detectors and traditional signature-based methods is conducted using a hand-curated dataset with both legitimate and malicious samples of emails. Criteria for evaluation include detection rates, false positives and negatives, as well as computational resources consumed. The experiments show that AI-based techniques outperform traditional systems in terms of recognizing unknown phishing emails. However, superior performance comes at the expense of greater computational loads and increased requirements for tuning and maintaining AI models. Also, this study provides practical guidance for integrating AI-based phishing detectors into corporate email systems, considering deployment issues, scaling, and computational resources needed. Based on the experiment results presented in the paper, recommendations are made regarding implementation of AI solutions for phishing attacks.

DOI: http://doi.org/10.5281/zenodo.20205928

Ai Prompt Helper

Authors: M.Benita Roy, L.Bhuvaneswar, K.V.Bharath Kumar, K.Vishnuvardhan

Abstract: The proliferation of digital content creation has created significant challenges for content creators, educators, and professionals who require efficient prompt engineering and AI-assisted content generation workflows. This paper proposes AI Prompt Helper, an intelligent system designed to optimize prompt construction, enhance content generation quality, and automate repetitive workflows for various AI applications. The proposed system integrates advanced natural language processing techniques, user-friendly interface design, and intelligent automation mechanisms to provide users with real-time suggestions, template management, and workflow optimization. The paper presents the system architecture, implementation details, performance evaluation, and practical use cases demonstrating the effectiveness of AI Prompt Helper in improving content generation productivity by up to 65% and reducing user cognitive load. Further, this paper provides insights into the system's design philosophy, technical implementation, and future research directions for intelligent prompt optimization systems.

AI-Powered Resume Analyzer & Smart Job Matching Platform

Authors: Vibhuti Chaddha, Nikunj Agarwal, Dr. Yatu Rani

Abstract: Abstract: The rapid growth of digital recruitment platforms and online hiring systems has significantly increased the volume of job applications received by organizations, making manual resume screening and candidate shortlisting time-consuming and inefficient. Traditional recruitment methods and keyword-based Applicant Tracking Systems (ATS) often fail to accurately identify suitable candidates due to limited semantic understanding and dependency on exact keyword matching. To address these limitations, this paper presents SmartHireX: AI-Powered Resume Analyzer & Smart Job Matching Platform, an intelligent recruitment system designed to automate resume analysis, improve candidate-job matching accuracy, and enhance recruitment efficiency. The platform is designed to support both recruiters and job seekers. Recruiters can automate candidate screening, ranking, and shortlisting processes, thereby reducing manual effort and improving hiring decision-making. Job seekers benefit from personalized resume improvement suggestions, missing keyword identification, skill gap analysis, and intelligent job recommendations that help optimize resumes according to industry and ATS standards.The system is implemented using modern web and AI technologies, including React and Tailwind CSS for frontend development, Flask (Python) for backend services, spaCy for Natural Language Processing tasks, and scikit-learn for Machine Learning and similarity analysis. Supabase is used for database management, while deployment can be supported using scalable cloud infrastructure. Experimental analysis and system evaluation demonstrate that the proposed platform improves recruitment accuracy, reduces screening time, and enhances candidate-job compatibility analysis compared to traditional recruitment approaches.

DOI: https://doi.org/10.5281/zenodo.20215014

Voice Controlled Wheel chair

Authors: Aayushree Chettri, Anup Ramudamu, Tripti Bhujel

Abstract: The paper illustrates the design and implementation of a basic speech reconnaissance wheelchair for alm civilian users. We layered a Bluetooth-capable Arduino Uno heart to create a system built around safety and security. We did this by taking the spoken intent and translating it into mechanical work to mitigate needing as much manual control over things, as well .We discuss the build process – the areasoning behind our obstacle-avoidance algorithms and what physically feasible hands-free mobility looks like.

Smart Wearable System for Continuous Soldier Location and Health Monitoring

Authors: Mr. Manvith M S, Ms. Varshitha A N, Ms. Varshitha C J, Mr. Vikas S, Dr. Nanda B S

Abstract: Smart Wearable System for Continuous Soldier Location and Health Monitoring is a smart, integrated solution designed to enhance soldier safety and mission efficiency. The system continuously measures vital signs such as heart rate, SpO₂, and body temperature using compact biomedical sensors. A GPS and IMU module provides accurate real- time location tracking, while communication ensures secure data transmission to the control center for monitoring and alerts. By combining vital monitoring, geolocation, and thermal regulation, the system supports early risk detection, improved soldier endurance, and better operational decision-making.

DOI: https://doi.org/10.5281/zenodo.20226278

Smart Water Purity Alert System

Authors: Ritesh Ashok Patil, Om Dilip Tambare, Ajay Madhav Tarte, Harshwardhan Pramod Bhosale, Mayur Maroti Pandhare, Ms. G.N. Mawale

Abstract: Access to clean drinking water is still a challenge for a large part of the world. Contamination from bacteria, dissolved solids, heavy metals, and chemicals continues to cause health problems, especially in areas where lab-based testing is slow or unavailable. This paper presents the Smart Water Purity Alert System (SWPAS), a low-cost device built around an ESP32 microcontroller with two water quality sensors measuring turbidity and TDS (total dissolved solids). Sensor data is sent to a cloud platform, and whenever any reading goes beyond the safe limit, users receive an alert via a mobile app, SMS, or a local buzzer. The system was tested on three water samples and correctly identified contaminated water without triggering false alarms on safe samples. Alerts were delivered within 2 to 4 seconds of a threshold breach, and the total hardware cost stayed under ■4,000.

DOI: http://doi.org/10.5281/zenodo.20226311

Hybrid Smart Mirror: A Multi-Modal IoT and AI Framework for Personalized Ambient Intelligence at the Edge

Authors: Rugved Daphal, Priya Karnawat, Harita Karnawat, Kadambari Nagare, Prof. Jameer Kotwal, Prof. Punam Raskar, Prof. Rucha Madali

Abstract: Smart mirrors represent chronically underutilized ambient interaction surfaces, yet existing implementations remain constrained by single-modality designs, absent user identification, insecure IoT integration, and inadequate empirical evaluation. This work introduces the Hybrid Smart Mirror (HSM) — a multi-modal ambient intelligence platform that fuses real-time biometric identification, natural language voice interaction, IoT device orchestration, and adaptive information rendering within a conventional mirror form factor, deployed entirely at the edge. A lightweight MobileNetV2 pipeline achieves 94.3% identification accuracy (F1 = 0.941) at 187 ms inference latency, with a 387 ms end-to-end pipeline from motion detection to personalized display. Voice command recognition achieves 6.8% Word Error Rate under controlled conditions; IoT commands are dispatched via TLS-encrypted MQTT at 43 ms round-trip latency. A STRIDE-informed security analysis underpins privacy-preserving countermeasures including on-device biometric storage and liveness detection. A proof-of-concept usability study (N=18) yields a SUS score of 82.4, exceeding the 68-point industry average and statistically outperforming five prior smart mirror systems (ANOVA F(3,68)=41.3, p<0.001). The HSM demonstrates that Raspberry Pi-class hardware can simultaneously achieve edge-native latency, sub-dollar-per-month operation, and GDPR-compliant privacy without sacrificing usability.

DOI: http://doi.org/10.5281/zenodo.20229647

Smart Career Recommendation System Powered By Artificial Intelligence For Optimized Career Planning And Decision-Making

Authors: Mr.Rajeshirke Satej Mahendra, Ms. Rajguru Jaya Mahendra, Mr.Sambherao Eshwar Mahadeo, Prof. V. A. Karad, Prof. S. B. Bhosale, Prof. A. P. Bangar, Dr. A. A. Khatri

Abstract: Traditional job portals rely on keyword matching, not allowing semantics of skills. Thus, they fail to connect candidates with jobs. Cursory manual resume screening takes 30-40 seconds. Conventional online assessments lack any integrity. DreamRole, an AI-based career recommendation and recruitment software which alleviates these essential limitations is presented in the paper. DreamRole uses Natural Language Processing (NLP) to parse resumes in PDF, DOC, DOCX, and TXT formats. It extracts skills, experience and education from a database of 200 technical skills. A matching algorithm based on the content generates a weighted relevance score that takes into consideration your skills (70%), experience (20%) and education (10%). It recommends only those positions that achieve a minimum threshold of 50 percent. The Semantic Skill Variation mapping helps in eliminating false negatives. By using OpenCV for real-time proctoring, the platform detects face abnormality, mobile phone presence, tab switching, forbidden key-stroke and terminates the test automatically on 3 successive occurrences of the same behaviour. A web speech API based hands-free navigation voice assistant (English, Hindi, Marathi) using a multilingual. Bulk resume processing can allow up to 10000 candidates in a single batch with automatic account generation feature and CSV credential download. The evaluation shows that we achieve an F1-score of 94.7% for parsing resumes, NDCG@10 of 0.79(27.4% improvement over the keyword baseline), proctoring precision of 95.7%. Moreover, our API responds under 700ms at 200 concurrent users. Finally, the evaluated System Usability Scale Score is 86.4, which indicates excellent usability.

DOI: http://doi.org/10.5281/zenodo.20230063

Real-Time Space Weather Prediction And Geomagnetic Storm Forecasting

Authors: Rajesh Kumar Mishra, Divyansh Mishra, Rekha Agarwal

Abstract: The coupling between solar activity, interplanetary magnetic field (IMF) dynamics, and Earth’s magnetosphere-ionosphere system gives rise to geomagnetic storms and substorms that represent a primary space weather hazard for technological infrastructure, including power grids, navigation systems, and low-Earth orbit (LEO) satellites. Traditional physics-based models rooted in magnetohydrodynamics (MHD) and empirical index formulations—such as the Dst and Kp indices—are physically interpretable but computationally expensive and insufficient for real-time forecasting at the 1–6 hour horizons required for operational applications. In this work, we develop and rigorously evaluate a hybrid Artificial Intelligence (AI) framework that fuses deep learning with physics-informed constraints to deliver accurate, uncertainty-quantified predictions of geomagnetic activity across multiple temporal horizons. Our ensemble approach achieves a root-mean-square error (RMSE) of 9.4 nT at 1-hour lead time and 18.7 nT at 6-hour lead time, representing improvements of 34% and 41%, respectively, over the best-performing operational empirical model (Temerin-Li). The full AI pipeline produces a complete geomagnetic forecast cycle in under 2 seconds on commodity GPU hardware, meeting real-time operational constraints.

Plastic Waste To Fuel Converter System: A Pyrolysis-Based Approach With Gas Filtration And Sensor Integration

Authors: Dr. Ashish Apate, Mr. Sandeep Patil, Mr. Paras Doshi, Ms. Sharvari Nikam, Ms.Tanishka Parjane, Mr. Prathmesh Chamalwad

Abstract: The increasing accumulation of plastic waste has become a major environmental concern due to its non-biodegradable nature and the limitations of traditional disposal practices. This study proposes a low-cost Plastic Waste to Fuel Converter System that transforms non-recyclable plastic into useful fuel through pyrolysis, a thermal decomposition process carried out in the absence of oxygen. The developed prototype incorporates an LPG-heated reactor, vapour condensation chamber, aluminium inline fuel filter, activated carbon gas purification unit, and MQ2 gas sensor for enhanced safety and emission control. Experimental testing using LDPE, HDPE, and PP plastic waste demonstrated successful production of combustible pyro-oil with characteristics similar to diesel-range fuel. The system provides an economical and environmentally conscious solution for decentralized waste management, particularly suitable for small communities and rural applications. The estimated prototype cost ranges from ₹2,860 to ₹4,150, supporting its feasibility for affordable implementation.

DOI: http://doi.org/10.5281/zenodo.20230605

Integrated Design Approach For Multi-Material Additive Manufacturing Of High-Strength Biocompatible Prosthetics

Authors: Ms. Aishvarya Thombare, Dr. Dilip Gangwani

Abstract: Additive manufacturing enables patient-specific prosthetic devices with complex geometries; however, single-material prints rarely meet concurrent requirements of stiffness, durability, and comfort. This study develops a low-similarity, citation-backed framework for multi-material 3D printed, load-bearing prosthetics. A region-wise material allocation strategy is proposed using PLA for structural shells, TPU for compliant interfaces, and carbon fiber–reinforced polymer (CFRP) for primary load paths. Finite Element Analysis (FEA) with validated boundary conditions (800–1500 N) and topology optimization are employed to minimize mass under stress constraints. Experimental validation (ASTM D638/D695) confirms a 31.8% increase in load capacity and a 23.6% reduction in deformation versus single-material PLA. The combined design–analysis workflow demonstrates improved fatigue resistance and user comfort while maintaining manufacturability.

DOI: http://doi.org/10.5281/zenodo.20268038

Fault Diagnosis Of Bearing Of Induction Motor Using CNN: A Review

Authors: Gaurav Suresh Ambatkar, Swapnil Choudhary

Abstract: Bearings are among the most failure-prone components in induction motors, and their degradation often leads to unplanned downtimes and considerable financial losses. Early and accurate fault diagnosis of motor bearings is thus a critical concern in industrial practice. While classical approaches such as vibration analysis and motor current signature analysis (MCSA) have long served this purpose, the emergence of convolutional neural networks (CNNs) has transformed the diagnosis process through automated feature extraction and superior multi-class discrimination. This paper presents a detailed and comprehensive review of recent advances in CNN-based bearing fault diagnosis for induction motors. Detailed types of bearing faults are discussed, alongside a critical comparison of traditional and CNN-based methodologies, the operational flow of CNNs, specific model architectures, and the standing of CNNs in relation to other leading diagnostic approaches

DOI: http://doi.org/10.5281/zenodo.20233165

Smart Industrial Safety Wearable Device Using Artificial Intelligence For Proactive Risk Prevention And Worker Protection

Authors: Mr. Sahil A. Bodke, Ms. Devika D. More, Ms. Samruddhi M. Pansare, Prof. P. A. Mande, Prof. A.P.Bangar, Prof. S. B. Bhosale

Abstract: Workers in industrial workplaces are still exposed to toxic gases and thermal stresses, are victims of mechanical injuries and have been.fatigue-related accidents. Conventional safety systems have remained reactive until now and have responded after an incident takes place. The emergence of Artificial Intelligence (AI), the Internet of Things (IoT), and cutting-edge wearable sensor technology is currently giving rise to new opportunities for proactive occupational safety. This paper presents a Smart Industrial Safety Wearable System (SISWS) whose performance is validated after prototype testing of around 1200 sensors observations under six hazards. The system shows 78% accuracy in hazard detection, 94% in PPE detection, 92% reliability in sensor performance, and can generate alerts in less than 3 seconds, contributing to a reduction of emergency response by 60%. The model will model safety conditions’ classification and predict risk using a hybrid Decision Tree and Long Short-Term Memory (LSTM). The selection of the model over Random Forest and pure CNN was driven by its aptness for edge deployment and its ability to identify temporal patterns in sequential sensor streams. The key research gaps identified in fatigue prediction in an industrial environment are: 1. Lack of multi-modal sensor fusion with real-time edge AI; 2. Insufficient datasets for industrial fatigue prediction; 3. Limited ergonomic wearables for a tough industrial environment; and 4. Lack of XAI in safety-critical decision-making. This study sets a solid base for further advancement involving AI to create an occupational safety system with a prevention focus.

DOI: http://doi.org/10.5281/zenodo.20233475

Soil Nutrient Prediction Model Using Data Mining Techniques For Sustainable Farming

Authors: Bhargavi M R, Anitha.V

Abstract: In precision agriculture, there is a need for precise, cost-efficient, and timely assessment of soil nutrients to ensure appropriate fertilization, minimize environmental risks, and maximize crop production. Laboratory soil analysis is considered highly accurate; however, it is relatively costly, laborious, and spatially limited. In this study, a holistic data mining model for predicting the content of soil macronutrients, including Nitrogen (N), Phosphorus (P), and Potassium (K), based on multiple soil samples collected from spatially distributed locations and multi-sensor fusion techniques, is described. The suggested model combines kriging interpolation, a Hybrid Random Forest-Multiple Linear Regression (RF-MLR) model with 73-87% accuracy, and ANN model with 89% accuracy for estimating N content. The effectiveness of the presented approach was tested for 2,500 soil samples collected from agricultural land, which showed that pH and EC values have a high correlation with the content of P and K, respectively, whereas OC level shows a significant correlation with the abundance of N. Overall, the developed method decreases testing expenses by 70% when compared to laboratory techniques, offering sufficient accuracy.

DOI: https://doi.org/10.5281/zenodo.20233719

AI-Driven Workforce Analytics For Predicting Employee Burnout, Engagement, And Retention In Hybrid Work Environments

Authors: Pushpendra Kumar Sharma, Dr.S.Sujatha

Abstract: The fast transition to hybrid work arrangements has dramatically changed employee relations, creating additional challenges for dealing with burnout, engagement, and retention. Survey-based solutions lack objectivity, bias-free data collection, and a forward-looking approach. We have developed a holistic AI solution for analyzing the workforces using various types of passive sensing data collected from collaboration software (Slack, email, Zoom), HR systems, and device usage records to forecast outcomes. The proposed MTL framework incorporates the attention mechanism and TCN to make predictions about three key metrics: burnout (classification task with 92% AUC), engagement (regression, with MAE equal to 0.34), and turnover risk (probability of leaving during the next 6 months with 89% AUC). Applied to 18 months of data from 5,000+ people at a major technology company, the model reveals behavioral features as key indicators: the presence of after-hours activity and the number of meetings show the highest correlations with burnout, whereas engagement is best predicted by peers' interaction diversity.

DOI: https://doi.org/10.5281/zenodo.20233792

Performance Evaluation Of BioBert And CNN Models For Neurological Disorder Detection

Authors: Abubakar Sadiq Muhammad, Salim Ahmad, Zaharaddeen S. Iro, Abba Dauda

Abstract: Accurate and efficient modeling of neurological disorders remains a significant challenge in clinical neuroscience. With the growing availability of unstructured clinical narratives, natural language processing (NLP) has emerged as a promising avenue for extracting diagnostic signals from text. This study presents a systematic comparison of two deep learning paradigms, transformer-based and convolution-based models for classifying neurological disorders from clinical notes. Specifically, we fine-tuned BioBERT, a domain-adapted transformer pretrained on biomedical corpora, and trained a Convolutional Neural Network (CNN) under identical experimental conditions, including dataset, preprocessing pipeline, hyperparameters (learning rate = 2e−5, batch size = 32, max length = 130), and evaluation metrics. BioBERT achieved 95.53% accuracy, 94.38% F1-score, and ROC-AUC of 0.952, significantly outperforming the CNN (89.62% accuracy, 88.32% F1-score, ROC-AUC = 0.918). The performance gap is attributed not to data or tuning advantages, but to fundamental differences in how the models process language: CNNs rely on local, n-gram–level pattern matching and fixed receptive fields, limiting their ability to resolve long-range dependencies and nuanced clinical expressions (e.g., negation, hedging, comorbidity descriptions); in contrast, BioBERT leverages bidirectional self-attention and domain-specific pretraining to capture contextual semantics, medical terminology, and subtle linguistic markers of pathology. These findings demonstrate that context-aware, domain-pretrained transformers offer a qualitatively distinct advantage over local-feature extractors like CNNs in clinical text understanding, supporting their integration into scalable, non-invasive diagnostic support systems.

DOI: https://doi.org/10.5281/zenodo.20234225

Microplastics As Modulators Of Soil Nutrient Cycling Under Hydrological Connectivity In Earth’s Critical Zone

Authors: Haider Ali, Hassan Ali, Komal Yaseen, Muhammad Ahmad, Waqas Altaf

Abstract: Microplastics have emerged as pervasive contaminants in terrestrial ecosystems, with agricultural soils serving as major sinks through sources including plastic mulch, sewage sludge, and atmospheric deposition. This comprehensive review synthesizes current knowledge on how microplastics modulate soil nutrient cycling within the Earth's Critical Zone, with particular emphasis on the mediating role of hydrological connectivity. The physical presence of microplastics fundamentally alters soil structure, porosity, aggregate stability, and hydraulic properties in polymer- and concentration-dependent manners, creating preferential flow pathways that govern both microplastic transport and nutrient dynamics. Biodegradable microplastics generally exert more pronounced effects than conventional polymers, increasing CO₂ emissions while simultaneously altering microbial carbon use efficiency. Nitrogen cycling is consistently disrupted, with reduced nitrate concentrations, enhanced N₂O emissions (up to 140.6% increase), and shifts in nitrogen-transforming microbial guilds. Tire wear particles and their associated chemicals, particularly the highly toxic transformation product 6PPD-quinone, represent emerging contaminants of high concern with substantial emissions (estimated 79.5 kt annually in the U.S.), accumulation in roadside soils (up to 26,400 mg/kg), and demonstrated toxicity to soil biota. The effects of microplastics on nutrient cycling are profoundly context-dependent, modulated by soil properties, microplastic characteristics, environmental conditions, and biological factors. Significant knowledge gaps remain regarding deep Critical Zone processes, temporal dynamics, and interactive effects with global change factors. As plastic production continues to increase, understanding and mitigating the effects of microplastics on soil nutrient cycling under varying hydrological regimes represents a pressing environmental imperative with implications for soil health, agricultural productivity, and climate regulation.

DOI: http://doi.org/10.5281/zenodo.20234336

Farmers’ Perceptions of Cooperative Marketing Society Effectiveness: A Multi-dimensional Analysis Using Factor and Cluster Analysis

Authors: Associate Professor Dr. S.Sureshbabu, Research Scholar Mr. A.kannan

Abstract: This study investigates farmers' perceptions of the effectiveness of marketing support provided by Cooperative Marketing Societies (CMS) through a comprehensive multi-dimensional analysis. Using exploratory factor analysis (EFA) and K-means cluster analysis on data from 620 farm-members, the research identifies two critical dimensions of marketing effectiveness: Cost Efficiency and Price Benefits, and Market Access and Sales Stability. The findings reveal that farmers strongly perceive CMS as effective in reducing transaction costs (mean = 3.76) and protecting them from middlemen exploitation (mean = 4.00), yet express relatively lower confidence in CMS's ability to expand sales volume (mean = 2.41) and ensure consistent market access (mean = 2.72). Cluster analysis reveals a tri-segmented farmer population: 65.2% highly satisfied, 18.9% moderately satisfied, and 16.0% less satisfied with CMS services. The study underscores the multifaceted nature of agricultural marketing effectiveness and provides actionable insights for CMS policy reformulation and targeted service improvements.

DOI: https://doi.org/10.5281/zenodo.20242461

Purchase Decision of Electric Vehicles: A Study with Special Reference to Coimbatore City

Authors: Dr.R.Nirmaladevi, K.Megalatha

Abstract: The global shift towards sustainable transportation has intensified interest in electric vehicles (EVs) as a viable alternative to conventional internal combustion engine (ICE) vehicles. India, as one of the fastest-growing automobile markets in the world, has witnessed a significant surge in EV adoption driven by government incentives, rising fuel prices, and heightened environmental awareness. Coimbatore, often referred to as the "Manchester of South India," is emerging as a key hub for EV adoption owing to its strong industrial base, educated urban population, and growing middle-class segment. This study aims to examine the factors influencing the purchase decision of electric vehicles among consumers in Coimbatore. A structured questionnaire was administered to a sample of 350 respondents selected through stratified random sampling. The data were analysed using descriptive statistics, chi-square tests, factor analysis, and multiple regression analysis. The findings reveal that government subsidies and incentives, total cost of ownership, environmental consciousness, technological features, charging infrastructure availability, and brand reputation are the primary determinants of EV purchase decisions. The study also identifies significant demographic differences in EV purchase intentions across gender, age, income, and educational qualification. The results provide actionable insights for automobile manufacturers, policymakers, and marketers to formulate effective strategies to accelerate EV adoption in Tier-II Indian cities.

DOI: https://doi.org/10.5281/zenodo.20242539

A Study On Secure Application Development Practices

Authors: Amartya Sahu Patel

Abstract: Secure application development practices are essential in today’s software-driven world, where applications are increasingly exposed to cyber threats, vulnerabilities, and data breaches. This study explores the principles, methodologies, and frameworks that ensure security is embedded throughout the software development lifecycle (SDLC). It highlights the importance of adopting a proactive approach through Secure Software Development Lifecycle (SSDLC), which integrates security at every phase, including requirements analysis, design, implementation, testing, deployment, and maintenance. The study examines key practices such as threat modeling, secure coding standards, code review, vulnerability assessment, and penetration testing. It also discusses the role of modern methodologies like DevSecOps, which integrates security into continuous integration and continuous deployment (CI/CD) pipelines. Furthermore, the paper addresses common security challenges such as injection attacks, authentication flaws, insecure APIs, and misconfigurations. Emerging solutions including automated security testing, static and dynamic analysis tools, and AI-driven vulnerability detection are also reviewed. The findings emphasize that implementing secure application development practices significantly reduces security risks, enhances software reliability, and protects sensitive data in modern digital environments.

DOI: http://doi.org/10.5281/zenodo.20283544

Distributed Data Processing Techniques In Cloud Systems

Authors: N. R. Rao

Abstract: Distributed data processing techniques in cloud systems have become fundamental for managing and analyzing large-scale data generated by modern applications. With the exponential growth of data from social media, IoT devices, enterprise systems, and web applications, traditional centralized processing approaches are no longer sufficient. Cloud-based distributed processing frameworks enable scalable, efficient, and fault-tolerant handling of massive datasets by distributing computational tasks across multiple nodes. This study explores key distributed processing models such as MapReduce, stream processing, batch processing, and in-memory computing. It also examines widely used frameworks including Hadoop, Spark, and Flink, highlighting their architectures and performance characteristics. The paper discusses how cloud environments support elasticity, parallelism, and high availability for large-scale data processing tasks. Additionally, it addresses challenges such as data consistency, network latency, fault tolerance, and resource optimization. Emerging trends such as serverless computing, edge-cloud collaboration, and real-time analytics are also reviewed. The findings emphasize that distributed data processing is essential for enabling efficient big data analytics, supporting scalable applications, and driving data-driven decision-making in cloud systems.

DOI: http://doi.org/10.5281/zenodo.20283506

Predicting Overtraining and Injury Risk in Marathon Athletes Using Machine Learning

Authors: Vikas Kumar, Dr. Akhilesh Das Gupta Institute of Professional Studies

Abstract: I built a machine learning model that can actually predict how much injury risk you have and what you might be doing wrong in your training. The idea came from my own personal experience — I was struggling with shin pain, tight hamstrings, calf tightness, thigh soreness, and sore knees after just my first week of running. I started watching YouTube videos and reading blogs to understand why this was happening. I learned about different running zones and why beginners who start too fast get injured early, how skipping warm-ups puts a cold body at serious injury risk, and how not eating enough or skipping recovery days makes everything worse. Basically injury does not come from one thing — it comes from getting multiple things wrong at the same time. Since collecting real athlete data from scratch would have taken years, I used a synthetically generated dataset of 1,240 weekly training records simulating 180 athletes, built around features like the Acute:Chronic Workload Ratio (ACWR), weekly mileage change, and recovery patterns. I tested four different classification models — Logistic Regression, Decision Tree, Random Forest, and XGBoost — to see which one would work best for this kind of prediction. XGBoost came out on top with 88.7% accuracy and ROC-AUC of 0.93 on the held-out test set. The model is deployed as a web app where the user fills in details like age, weekly mileage, height, mileage change percentage, recovery days, and active days — and gets back their injury risk percentage along with suggestions to help them stay injury-free.

DOI: https://doi.org/10.5281/zenodo.20266328

Smart Crop Advisory System Using IoT and AI

Authors: Sukanya Suresh Salunke, Austin Frank, Aditya Bhosale, Samaira Bhandkar, Sayali Jadhav, Mr. N. D. Gaikwad, Prof.N.D Gaikwad

Abstract: Nowadays, with the rapid urbanisation and industrialisation in the recent decades, the need of effective environmental monitoring has increased for sustainable development and public health. The work presented here is the design and implementation of an intelligent Internet of Things (IoT) environmental monitoring system, which is intended to provide real-time, accurate and scalable solutions for the acquisition and analysis of environmental data. The proposed system combines state-of-the-art sensor networks with embedded intelligence to monitor important environmental parameters such as air quality, temperature, humidity and noise levels. Remote monitoring and centralised control are achieved through the use of wireless communication protocols and cloud-based data management, which allows for early detection of environmental anomalies and trends. The main contribution of this paper is the integration of machine learning algorithms for predictive analytics and anomaly detection, improving the system’s responsiveness and reliability. The modular architecture allows for easy integration of new sensors and capabilities and provides flexibility for different deployment scenarios, from urban centres to industrial zones and rural areas. Extensive field experiments and performance evaluations have demonstrated the system's effectiveness in providing high-resolution environmental insights at low power consumption and with little maintenance. This research indicates the transformative potential of intelligent IoT systems in environmental monitoring, providing a scalable and cost-effective framework for policy makers, environmental agencies and researchers. Results pave the way for future work on smart environmental management and contribute to the broader vision of sustainable, data-driven urban ecosystems.

DOI: https://doi.org/10.5281/zenodo.20265552

Analysis Of Modularity Based Community Detection Algorithms To Detect Communities In Social Network Analysis

Authors: Deepali Piple, Mukesh Sakle, Shaligram Prajapat

Abstract: Social Network Analysis relies upon community detection to identify groups of nodes within a network that maintain stronger inter-node relationships than their node linkages to all other components of the network. Community Detection enables researchers to understand how complex networks construct and establish their functional connections. This research tests various community detection techniques by way of several modularity-based community detection techniques using established benchmark social networks. The study evaluates four community detection algorithms: Label Propagation Algorithm (LPA) and Clauset-Newman-Moore (CNM) and Louvain and Leiden.Each of the four community detection algorithms was evaluated against five standard benchmark social networks including Karate, Dolphin, Football, Facebook, Polbooks, Les Misérables and Jazz.Performance evaluation metrics used for each of the four algorithms included the Modularity Index (MI), Conductance, Normalized Mutual Information (NMI) and the Adjusted Rand Index (ARI). Results from this research demonstrated that the Louvain method consistently produced higher MI values and had consistent performance across all of the various test data sets, while the LPA method performed all other methods in networks where community structures are clearly visible. The research also found that modularity-based optimizations successfully identified critical social network communities.

DOI: http://doi.org/10.5281/zenodo.20265823

An Android-Based Tourism Navigation And Information System For Uganda

Authors: Joseph Kiggundu, Dr. Vaibhav Bhushan Tyagi

Abstract: This paper explores the tourism sector which has experienced a rapid digital transformation due to the increasing use of mobile technologies and online platforms. However, in many local tourism environments, interactions between the tourist and the guides remains largely informal, often relying on personal contacts and unstructured communication channels. This lack of a structured digital platform creates challenges in service discovery, booking coordination, transparency, and accountability among tour guides. This project presents the design and implementation of a Local Tourism Guide Android Application aimed at improving interaction between tourists and registered local tour guides. The system introduces a centralized mobile platform that allows tourists to search for available guides, request bookings, negotiated service quotations, and communicate with guides through a real-time messaging feature. In addition, the system incorporates a guide performance evaluation mechanism based on booking frequency, ratings, and service activity to promote transparency and service quality. The system is implemented using Flutter for the mobile application interface, PHP for the backend RESTful API services, and MySQL as the database management system. Integration with location services enables tourists to discover guides based on geographical proximity. An administrative module is also included to support system governance through user management, booking monitoring, and performance analytics. The developed application aims to enhance coordination within the local tourism ecosystem by introducing structured workflows for guide discovery, booking management, communication, and service evaluation. By providing a unified digital platform, the system seeks to improve the reliability, accessibility, and transparency of local tourism services.

DOI: https://doi.org/10.5281/zenodo.20265958

Domain-Based Career Roadmap Management System With Integrated AI-Driven YouTube Channel Recommendation (DCRMS)

Authors: G. Vedavyas, G. Hareesh, M. Ajith Kumar Reddy

Abstract: The proliferation of online educational content has deepened learner disorientation rather than alleviating it. Despite millions of instructional resources available across platforms such as YouTube, Coursera, and Khan Academy, the absence of a unified, domain-aware guidance mechanism forces learners into fragmented self-directed journeys. This paper presents the DomainBased Career Roadmap Management System (DCRMS), a webbased educational technology platform that integrates structured roadmap generation, AI-assisted personalized recommendation, NLP-based skill extraction, and learning progress tracking into a single coherent interface. The proposed architecture combines a rule-based domain engine with a hybrid recommendation pipeline that uses collaborative filtering, content-based similarity scoring, and a Directed Acyclic Graph (DAG) traversal model to construct prerequisite-aware learning paths. A Skill-Gap Computation Module employs cosine similarity between user-skill vectors and domain competency matrices to identify and prioritize learning deficiencies from uploaded resumes. The prototype recommendation engine achieves a Precision@10 of 0.847, Recall@10 of 0.791, and an F1-score of 0.818 across seven IT career domains in experimental evaluation. Under simulated load of 50 concurrent users, roadmap generation latency averages 318 ms. A structured usability study with 50 undergraduate participants yields a System Usability Scale (SUS) score of 82.4 (Grade B+), confirming strong user acceptance. DCRMS constitutes a reproducible proof-of-concept contribution to educational data mining, career guidance, and intelligent recommender systems for the informal learning context.

AI-Powered Interview Automation System For Real-Time Candidate Assessment

Authors: Ravanaboyine Saisri, N.Sandhya Rani

Abstract: Automated intelligent software agents may mimic human communication behaviours, allowing for more organic and interesting interactions with people, thanks to the fast development of conversational AI. With these new developments, it's possible to replace human interviewers with autonomous software agents that are smarter than humans, thereby automating the candidate interview process. Conversational AI allows machines to mimic human interviewers in many ways, including asking questions, understanding and evaluating responses, and starting dynamic discussions. Improving the efficiency of the whole interview process, this automation guarantees consistent and impartial assessment, which in turn leads to more effective and fair recruiting procedures. An AI-driven interview system that can evaluate candidates in real time is the focus of this research article, which intends to provide a thorough analysis of its design and implementation. Various artificial intelligence agents work together in this system to do things like choose questions based on predefined criteria, evaluate candidates' responses, analyse speech for signs of emotion and sentiment, and then combine these findings to give distinct scores for answers and emotions in the performance evaluation. Index Terms—Generative Pre-trained Transformers (GPT), AI-driven interviews, automated candidate assessments, multimodal emotion detection, and natural language processing (NLP).

Crop Cart-An Implementation Where Farms Meet Families

Authors: Shrija R.Pansare, Srushti S.Pawar, Ankita S.Kunjir

Abstract: CropCart Management System is a digital platform designed to connect farmers directly with consumers (families) without the involvement of middlemen. In the traditional agricultural system, farmers often receive low prices for their produce, while consumers end up paying higher prices. This happens due to the long supply chain and lack of direct communication between farmers and buyers.The main aim of CropCart is to create a simple and efficient system where farmers can upload their products, set prices, and sell directly to customers. On the other hand, consumers can browse fresh fruits, vegetables, and other farm products, and purchase them easily from nearby farms. This ensures better profit for farmers and fresh, affordable products for customers. The system uses basic digital technologies to manage product listings, orders, and user details. It also improves transparency by providing information about the source of products, which builds trust among users. By reducing intermediaries and simplifying the supply chain, CropCart helps in minimizing wastage and improving efficiency.Overall, CropCart Management System is a step towards smart agriculture and digital transformation, making the farming business more sustainable and beneficial for both farmers and consumers. Keywords: Crop Management, Farm Records, Crop Monitoring, Sustainable Farming, Yield Analysis, Agri-Tech.

Effect of Different Types of Infills on The Static and Dynamic Behaviour of Rc Frame Structures With And Without Retrofitting

Authors: Shreyanshu verma

Abstract: This study investigates the effect of different types of infill walls on the static and dynamic behavior of reinforced concrete (RC) framed structures with and without retrofitting. Masonry infill walls significantly influence the stiffness, strength, and seismic performance of RC frames, although they are often neglected during structural analysis. In this work, bare frames and infilled frames using conventional brick masonry and AAC blocks are analyzed under seismic loading using equivalent static and time history methods according to IS 1893:2016 provisions. Numerical modeling and finite element analysis are carried out using ABAQUS software. The study also evaluates the effect of soft stories at different floor levels and the efficiency of CFRP retrofitting techniques. Results indicate that infilled frames show considerable reduction in displacement and improved stiffness compared to bare frames. AAC infill walls provide better seismic performance than conventional masonry. Retrofitting using CFRP further enhances structural resistance and reduces seismic displacement significantly.

DOI: http://doi.org/10.5281/zenodo.20267094

Integrated Design Approach For Multi-Material Additive Manufacturing Of High-Strength Biocompatible Prosthetics

Authors: Ms. Aishvarya Thombare, Dr. Dilip Gangwani

Abstract: Additive manufacturing enables patient-specific prosthetic devices with complex geometries; however, single-material prints rarely meet concurrent requirements of stiffness, durability, and comfort. This study develops a low-similarity, citation-backed framework for multi-material 3D printed, load-bearing prosthetics. A region-wise material allocation strategy is proposed using PLA for structural shells, TPU for compliant interfaces, and carbon fiber–reinforced polymer (CFRP) for primary load paths. Finite Element Analysis (FEA) with validated boundary conditions (800–1500 N) and topology optimization are employed to minimize mass under stress constraints. Experimental validation (ASTM D638/D695) confirms a 31.8% increase in load capacity and a 23.6% reduction in deformation versus single-material PLA. The combined design–analysis workflow demonstrates improved fatigue resistance and user comfort while maintaining manufacturability.

DOI: http://doi.org/10.5281/zenodo.20268038

Optimization Of Activated Sludge Process For Enhanced Nitrogen And Phosphorus Removal: Integrating The ASM2d With Floc Model And SRT Control

Authors: Aboajela Musbah Kajaman

Abstract: To reduce energy consumption, and to achieve the desired denitrification, the activated sludge process sometimes needs to operate at low dissolved oxygen concentrations. The ASM2d model describes the activated sludge process during different phases in a sequencing batch reactor. A comprehensive floc model remains lacking despite the widespread study of enhanced biological phosphorus removal. As a result, the integrated system model is developed to understand the impact of floc at low DO concentrations, during biological nitrogen and phosphorus removal. Additionally, optimisation of parameters and effectiveness factors for developing of the ASM2d with floc model could be achieved by controlling sludge retention time (SRT), particularly during the activated sludge process. In a wastewater treatment plant, the dissolved oxygen was controlled at a low concentration, and its dispersion coefficient into the floc is 〖1.2×10〗^(-4) m^2/day. This study aims to optimise the activated sludge process, which is critical for efficient wastewater treatment by controlling the SRT and the amount of sludge access. The objective is to predict process behaviour under various operational configurations, based on differing sludge wasting times.

DOI: https://doi.org/10.5281/zenodo.20268292

Automated And Fully Flexible Hall Allotment System

Authors: Assistant Professer.Ms.B. Sasi, Y.V. Karthik Reddy, V. Simhadri, S. Bharath Kumar

Abstract: The Evolution of a college test Hall Seating Arrangement System enhanced the efficiency of test hall allocation and student seating arrangements. The System allows easy retrieval of examination information for a student who is in a particular class. These systems are intended to automate the manual procedures in order to make the examination administration processes more efficient. One of the reasons it was developed is that the system is Proficient of creating the report seating arrangement automatically during or after the session and at breaks during the examination. The project’s scope is the system on which the program is installed. In other words, the project is intended to be a web application that is accessible to a particular organization. Most students at the college face numerous challenges with finding the examination room and their designated seats within. A new concept that has been implemented allows students to monitor controlled access to their examination sessions. This feature allows them to obtain advanced floor or directional guidance to the designated halls. In the Students' Information, we get records of all the candidates who sit for the examination. It contains the student's name, hall ticket number, branch, and hall number. Hall Details contains each hall name and their corresponding total number of halls available in the institution. The information of the examination schedules includes the total time allocated to the student and the hall among other things. The batch details incorporate department details like computer science, biology, chemistry, mathematics, etc. The project monitors several details in modules such as student information and exam timing information.

Artificial Intelligence Based Predictive Modeling For Smart Decision Support Systems

Authors: Shah Md. Tanzimul Kabir, Md. Yusuf Miah

Abstract: This paper provides a comprehensive analysis of the predictive modeling frameworks of Artificial Intelligence for smart decision support systems (DSS) and the evolution of the traditional analytics approach to an integrated Artificial Intelligence approach for real-time decision-making. Through a systematic study of the recent research articles from 2021 to 2026, the paper explores the advancements in machine learning models and hybrid Artificial Intelligence approaches for transforming the traditional decision support systems in the healthcare, finance, manufacturing, and environmental management domains. The research proposes an Integrated Predictive Decision Support Framework (IPDSF) that incorporates data preprocessing, model selection, explainability, and human validation for effective predictions. The study reveals that the contemporary Artificial Intelligence-based decision support systems employ ensemble learning (Random Forest and XGBoost with an accuracy rate of 89-96%), deep learning for complex pattern recognition (CNN for medical image analysis and LSTM for time series analysis), and hybrid neuro-symbolic models for effective predictions. Some challenges still exist in model interpretability, but Explainable AI (XAI) techniques such as SHAP and LIME have become a key component in building user trust and ensuring compliance. Comparative evaluation of AI-DSS along four analytical dimensions—model architecture, interpretability, real-time, and domain adaptation—clearly shows that an appropriate balance between predictive accuracy, interpretability, efficiency, and integration with existing decision processes is required.

DOI: https://doi.org/10.5281/zenodo.20270472

Online Signature Verification Using Siamese Convolutional Neural Networks For Secure Digital Authentication

Authors: Shaik Nakarikanti Shabana, P. Anusha

Abstract: Secure authentication in digital settings that electronically collect and verify handwritten signatures relies heavily on Online Signature Verification (OSV) technologies. These systems analyse not only the static structure of a signature but also its unique dynamic properties like pen pressure, pace, and stroke order using machine learning and deep learning methodologies. Hybrid models, Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), and Generator Adversarial Networks (GANs) are some of the current methods used to examine the distinct static and dynamic aspects of signatures. The above deep learning approaches provide a challenge when it comes to developing responsive and dependable real-time systems. This is because these models need to be trained every time a new user is added to the database. Our experiment's goal is to construct an OSV system that comprises a ReactJS-built website for signature uploading or database storage and a CNN-based Siamese Network for OSV integration. The model uses the spatial information extracted from the signatures to make judgements depending on how close the uploaded signature is to the user's original signature. Through ongoing model training, the system is prepared to deal with both genuine and fake signatures. This system is designed to provide a safe, dependable, and resilient way to verify identification in online transactions and sensitive digital applications. It does this by combining signature preprocessing methods, feature extraction, and classification models. Things like online signature verification, siamese networks, react-js, and feature extraction are all part of the index.

DOI: http://doi.org/10.5281/zenodo.20271121

Modern Restaurant Management System

Authors: Boddu Mahendra Reddy, Damireddy Ranga Reddy, Chagam Anvesh Reddy, Benita Roy

Abstract: A Modern Restaurant Management System using Java and MySQL is a software application developed to automate and manage important restaurant operations such as order processing, billing, inventory management, employee management, and table reservation efficiently. The system is designed using object-oriented programming concepts in Java, which provide modularity, reliability, scalability, and easy maintenance. The application includes different modules for user authentication, menu management, customer order handling, automated billing, and inventory monitoring.The system uses MySQL as the backend database and JDBC for database connectivity, enabling secure storage, retrieval, and management of restaurant data. It supports real-time order processing, automatic bill generation with tax calculations, sales record maintenance, and stock updates whenever orders are processed. The application helps restaurant staff perform daily activities more accurately while reducing manual effort and operational errors.The proposed system improves restaurant productivity, reduces paperwork, enhances customer service quality, and provides better control over restaurant operations. It also generates useful reports related to sales, inventory, employee records, and customer orders, making restaurant management more efficient, organized, and reliable.

DOI: http://doi.org/10.5281/zenodo.20280251

A Trust-Aware Blockchain And AI-Driven Secure Communication Framework For Internet Of Vehicles Using Hybrid SDN And Modified PBFT

Authors: Dr.Ritu Agarwal, Aryan Sharma, Abhinav Singh

Abstract: The Internet of Vehicles (IoV) is playing a critical role in modern transportation systems, as vehicles interact with each other and roadside infrastructures in real-time. Although this technology can increase road safety and traffic efficiency, it creates various security challenges for IoV. For instance, Sybil, replay, and impersonation attacks can be launched on IoV networks, thereby making them susceptible to attacks. In this paper, we propose a hybrid model that utilizes blockchain tech-nology, SDN, and artificial intelligence-based anomaly detection mechanisms. Specifically, we use random forest to detect abnor-mal behavior of malicious vehicles based on their interactions. Moreover, a modified PBFT (mPBFT) algorithm is employed to improve consensus efficiency by selecting reliable nodes. The proposed model is tested under a simulated IoV envi-ronment and obtains approximately 92detection. In addition, it significantly decreases consensus delay as compared to the conventional PBFT algorithm. Overall, this model has shown good efficiency, scalability, and security in IoV systems.

DOI: http://doi.org/10.5281/zenodo.20280734

An Intelligent Healthcare Prediction System Using Ensemble Machine Learning And Generative AI

Authors: Pujala Sivaleela, M.Prasanna Kumar

Abstract: Today, because to advancements in both technology and healthcare, people may have access to affordable, personalised health assistance right where they are. This research presents a method that may help individuals with their health by giving them detailed information on their health and being able to forecast when they may become sick. In order to foretell potential health problems, users are asked to provide three main symptoms together with basic biographical details such as gender and age. Decision trees, random forests, naive bayes, logistic regression, support vector machines (SVMs), K-nearest neighbours, and XGBoost are the seven machine learning models that the system uses to provide a final prediction using a voting ensemble model. In order to provide thorough descriptions of the expected illness, including its origins, symptoms, potential treatments, and remedies, the system integrates generative AI with predictive analytics. The generative AI model takes into account the user's age and gender to provide information that is personalised to their requirements. The reliability and effectiveness of the system may be shown by evaluation metrics including F1-score, accuracy, precision, and recall. People are better able to make informed choices about their health because to our user-friendly platform that makes health information freely accessible.

DOI: http://doi.org/10.5281/zenodo.20280806

AI-Driven Drug Sensitivity Prediction Using COSMIC, DGIdb, And GDSC Integration

Authors: Gurram Lavanya, Ch.Naveen

Abstract: It is critical to investigate the link between treatment efficacy and sensitivity to mutational patterns in order to successfully treat complex diseases such as cancer. Specifically, cancer drugs may lose some of their efficacy over time as a result of new mutations, and cancer cells themselves are continually changing as a result of ongoing mutations. The main purpose of this research is to analyse the relationship between medications, disorders, and genes using statistical methods. To further anticipate the drug sensitivity of cancer cells based on genomic changes, a generic processing pipeline and machine learning models were developed. This was accomplished by comparing an improved database to four well-known open-source databases: one that had information on drug sensitivity in cancer cell lines, one that contained resources for somatic mutation data, and the other that contained information on gene-drug interactions. Text encoding, filtering, and optimisation were among the preprocessing procedures used to provide a fresh, enhanced dataset for use in machine learning and statistical analysis. Statistics were run on the supplemented database to find out how much of an impact gene-drug interactions had on drug sensitivity. On the other hand, machine learning algorithms that have been trained on datasets of drug interactions or somatic mutations may be used to forecast drug sensitivity. Incorporating feature significance and ablation studies, the research aims to provide a thorough analysis of gene and pharmaceutical sensitivity. A constant R2 score of 0.73 across several data sources, on top of an R2 score of 0.91 in early testing, demonstrated good generalizability for the built pipeline. Ablation studies, statistical analysis, and machine learning all provide new perspectives to the field of drug sensitivity prediction.

DOI: http://doi.org/10.5281/zenodo.20281858

Evaluating ERP System Efficiency In The Indian Apparel Retail Industry: An Exploratory Study Using A Mixed Method Approach”

Authors: Dr. Arpita Pandey, Abhishek Ananthanarayanan

Abstract: The Indian retail sector is marked by high product complexity, volatile demand patterns, and intense competition, necessitating robust digital systems for operational integration and efficiency. Enterprise Resource Planning (ERP) systems are widely adopted across retailers; however, empirical evidence on their actual efficiency outcomes in the Indian retail context remains limited. This study investigates the effectiveness of ERP systems in enhancing operational efficiency and organizational performance in Indian retail firms, using a mixed-method approach. Primary data were collected from retail professionals through a structured questionnaire, supplemented by qualitative insights. Advanced analytical techniques—Partial Least Squares Structural Equation Modelling (PLS-SEM) and Analytic Hierarchy Process (AHP)—were employed to strengthen analytical rigor. The SEM results reveal that ERP implementation effectiveness has a significant positive impact on operational efficiency, which in turn strongly influences organizational performance, indicating a partial mediation effect. AHP findings prioritize inventory management and supply chain management as the most critical ERP modules driving efficiency in retail. The study contributes to the ERP and retail management literature by providing India-specific empirical evidence and a dual-method analytical framework. From a managerial perspective, the findings highlight the importance of strategic ERP module prioritization, continuous employee training, and change management to maximize ERP returns. The study offers actionable insights for retail managers, ERP vendors, and policymakers involved in the digital transformation of the sector.

DOI: http://doi.org/10.5281/zenodo.20284571

Medical Image Analysis With AI Assist

Authors: Mynam Hemanth Kumar, Koparthi Mahendra, Koppala Vijay Raju, Ms.G.Archana

Abstract: The proliferation of digital medical imaging has created an urgent need for automated, accessible tools capable of performing image segmentation and clinical interpretation without requiring specialist infrastructure. This paper presents Medical Image Analysis with AI Assist, a web-based medical imaging platform developed in Python and Streamlit. The system integrates a dual-pathway segmentation architecture — a primary UNet/DOSMA pipeline for DICOM/NIfTI data and a CLAHE-Otsu-morphological fallback for standard formats — with pixel-level statistical analysis, Google Gemini 1.5 Flash-powered clinical summarisation, and a stateful conversational chatbot. Evaluation on 150 de-identified images across five modalities yields: segmentation coverage accuracy 91.4%, expert-rated AI summary quality 4.2/5.0 (Krippendorff's α = 0.74, three raters), chatbot relevance 4.2/5.0, and mean processing time 16.8 s. CLAHE parameters (clip limit 2.5, 8×8 tile) are empirically justified; evaluation methodology, report completeness measurement, comparative fairness, patient privacy, data governance, clinical safety, hardware specification, and state-of-the-art benchmarking are explicitly addressed throughout.

Rewards As Reinforcement And Intervention To Enhance Task Responses Of A Learner With Autism

Authors: Ababon, Kyla Donna G., Banguis,Kelsie Camile, Susan Vincent D. Villarente

Abstract: This research aimed to use rewards as intervention in enhancing the performance of a Self-contained learner with Autism since the problem of the learner observed by the researchers are taking a long time to respond and a lot of pauses in between tasks. The participant of the study is an Autism Learner in a self-contained class. The researchers used rewards like toys, snacks and break time as an intervention in enhancing the performance of the learner and minimize the pauses between tasks. The researchers have conducted a pre-intervention observation assessment the result shows a mean score of 1 followed by a post-intervention observation assessment with a mean of 3. This means that the use of rewards like toys, snacks and break time as an intervention have enhanced the performance of the learner and minimizes pauses between tasks.

Design and Fabrication Of Head Motion Control Based Wheel Chair

Authors: Mr. Kushal Gowda U, Mr. Linith R, Mr. Lokesh M E, Mr. Madhu K R, Mr. Syed Imran Ali Asst. Professor

Abstract: This paper presents the design, fabrication, and experimental analysis of a Head Motion Control Based Wheelchair for individuals with severe motor disabilities. The system utilises an MPU-6050 three-axis accelerometer and gyroscope module mounted on the user's head to detect intentional head tilts — forward, backward, left, right, and neutral — and translates them into corresponding wheelchair drive commands via an ESP32 microcontroller and L298N motor driver. Experimental results across 300 trials yield an overall gesture recognition accuracy of 97.3%, average system latency of 124 ms, and successful multi-directional navigation in indoor environments. The system outperforms all prior comparable implementations in accuracy, latency, and is the only design to incorporate continuously variable speed control proportional to tilt angle magnitude

DOI: http://doi.org/10.5281/zenodo.20286178

Plastic Waste to Pyrolysis Oil Conversion

Authors: Mr.Chethan Kumar B M, Mr. Chiranjeevi B K, Mr.Darshan B R, Dr. Mohammad Rafi. H. Kerur, Proff. Somashekar B R

Abstract: This project presents the design and development of a Plastic Waste to Pyrolysis Conversion system — an innovative waste-to-energy solution. Plastic waste has become one of the most critical environmental challenges due to its non-biodegradable nature and widespread usage. The project focuses on converting plastic waste into pyrolysis oil through controlled thermal degradation in an oxygen-free environment. The obtained crude oil is further refined through fractional distillation and upgrading techniques to yield diesel-range fuel. Key processes include catalytic cracking and hydrotreatment to improve fuel quality. Testing of physical and chemical properties such as calorific value, viscosity, flash point, and density confirms the viability of the produced fuel as a sustainable alternative to conventional diesel.

Leafscan Ai: Deep Learning-Based Plant Leaf Disease Detection System

Authors: Kartikkey Uttam Khot, Srikant Rajkumar Kadam, Prathmesh Vijay Pawa, Pranit Chandrakant Gaikwad

Abstract: Plant diseases are one of the leading causes of agricultural productivity loss worldwide, threatening food security and farmer livelihoods. Traditional disease identification methods rely on manual visual inspection by agricultural experts, a process that is slow, expensive, and prone to error. This paper presents LeafScan AI, a deep learning-based plant leaf disease detection system capable of identifying 33 distinct disease categories across 9 plant species — Apple, Cherry, Corn, Grape, Peach, Pepper, Potato, Strawberry, and Tomato — with up to 96.88% classification accuracy. The system employs a Convolutional Neural Network (CNN) trained on the publicly available PlantVillage dataset, integrated with a Flask web application for real-time prediction via image upload. OpenCV is used for image preprocessing, and TensorFlow/Keras provides the deep learning backbone. The proposed system enables farmers and agricultural professionals to obtain instant, accurate disease diagnoses from leaf photographs, facilitating timely intervention and minimizing crop losses. Experimental results demonstrate high accuracy across diverse disease categories, validating the practical deployability of the model for smart agriculture applications.

DOI: https://doi.org/10.5281/zenodo.20302936

Heart Disease Prediction Using Logistics Regression

Authors: Murari Raja Krishna, Maddisetti Balu, Muppalla Ravi, Ms. Sivasangari

Abstract: Abstract- Cardiovascular disease remains the leading cause of global mortality, claiming approximately 17.9 million lives annually. Early and accurate prediction of individual patient risk is essential for enabling timely clinical intervention and reducing preventable mortality. This paper presents a comprehensive machine learning–based clinical decision support system for heart disease prediction using Logistic Regression, deployed as an interactive Streamlit web application. The system is trained and evaluated on the UCI Cleveland Heart Disease dataset comprising 303 patient records and 13 clinical attributes. The Logistic Regression model achieves a test accuracy of 88.52%, F1-score of 0.89, and AUC-ROC of 0.9267 on a held-out 20% test partition. Rigorous comparative evaluation against five baseline classifiers—Naïve Bayes, K-Nearest Neighbors, Support Vector Machine, Decision Tree, and Random Forest—confirms that Logistic Regression provides the best overall balance of interpretability, predictive performance, and generalization stability. The Streamlit web application enables real-time risk stratification and probabilistic output visualization across all 13 clinical features, bridging the gap between algorithmic performance and practical clinical utility.

Clause Guard : An app that lets users to detect loopholes in terms and conditions.

Authors: Jaya Veeranjaneya Reddy B, D Nikhil, Ch Rajesh, Mr. C Ram Chandran

Abstract: In the rapidly evolving digital ecosystem, users frequently interact with mobile applications, websites, e-commerce platforms, and online services that require acceptance of lengthy Terms and Conditions (T&C), Privacy Policies, and User Agreements. Most users tend to accept these agreements without thoroughly reading or understanding the hidden clauses, legal complexities, data-sharing permissions, subscription traps, liability waivers, and privacy loopholes embedded within them. This creates a significant gap between user awareness and digital consent, often leading to privacy violations, unauthorized data collection, financial exploitation, auto-renewal scams, and misuse of personal information. To address this growing concern, Clause Guard is proposed as an intelligent and user-centric application designed to detect, analyze, and simplify loopholes present in legal agreements of apps and websites. Clause Guard leverages advanced technologies such as Artificial Intelligence (AI), Natural Language Processing (NLP), Machine Learning (ML), and Legal Text Analytics to automatically scan and interpret complex legal documents in real time. The application identifies potentially harmful, suspicious, or manipulative clauses and categorizes them based on risk level, transparency, and user impact. It highlights critical sections related to data privacy, third-party data sharing, hidden charges, subscription auto-renewals, liability disclaimers, intellectual property rights, tracking permissions, and consent manipulation practices. By converting complicated legal terminology into simple and understandable language, the platform enhances digital literacy and empowers users to make informed decisions before accepting agreements.

Quantum Machine Learning for Advanced Climate Pattern Detection Using Satellite Data

Authors: Mrs K. Rajavadhani, Jerlin Flowrence D, Kasturi Uday Kiran, Challagundla Rakesh

Abstract: Climate pattern detection through satellite remote sensing is a critical task for understanding environmental changes and anomalies. However, current systems do not have an interactive interface for efficient visualisation and analysis of complex climate data. This study aims to develop a hybrid Quantum Machine Learning (QML) approach for sophisticated climate pattern detection through satellite data. The proposed system combines Principal Component Analysis (PCA) for dimensionality reduction with quantum-classification algorithms, such as the Variational Quantum Classifier (VQC) and the Quantum Support Vector Machine (QSVM), to detect complex climate patterns. A Streamlit web interface is designed to offer an interactive platform for data entry, visualisation, and monitoring. The system design includes a frontend interface unit, a data integration unit, and a planned AI & control unit for the backend quantum processing system. At present, the system design includes the implementation of frontend visualisation and user input modules using sample, publicly available satellite climate data. The integration of the quantum backend and direct satellite data connectivity will be done in future stages. The proposed method aims to leverage classical data preprocessing and quantum machine learning for improved anomaly detection and interpretability of satellite climate data.

AI-Based Job Skill Analyzer

Authors: Saragadam Tarun, M V Pavan Kumar, Potu Srikanth, Ms R Santhana Lakshmi

Abstract: The rapid advancement of digital recruitment platforms and the exponential growth in job applications have significantly increased the complexity of candidate screening and recruitment management. Traditional recruitment systems largely depend on manual resume screening, which is time-consuming, inefficient, and often unable to accurately identify the most suitable candidates for specific job roles. Furthermore, job seekers frequently struggle to understand industry requirements, optimize their resumes, and identify the skill gaps preventing them from securing suitable opportunities. To address these challenges, this paper proposes an AI-Based Job Skill Analyzer that integrates Natural Language Processing (NLP), Machine Learning, and Web Automation technologies to automate and optimize recruitment workflows. The proposed system extract candidate information from resumes in PDF format, identifies technical and professional skills using NLP algorithms, scrapes real-time job listings from multiple recruitment platforms, and performs intelligent resume-job matching using TF-IDF vectorization and Cosine Similarity algorithms. The system further performs skill gap analysis by comparing candidate competencies with industry requirements and provides personalized recommendations for improvement. A dedicated recruiter dashboard enables recruiters to filter candidates based on skills, match scores, salary expectations, and experience levels, thereby improving recruitment efficiency and decision-making processes. The implementation of this system demonstrates significant improvements in recruitment automation, candidate-job alignment, and hiring accuracy while reducing manual effort and operational costs. The proposed architecture is scalable, secure, and adaptable to future AI-driven recruitment ecosystems.

An AI-Driven Budget Revenue Projection System For Public Finance: Architecture, Data Model And Implementation

Authors: Dr. Bayomock Linwa Andre Claude, Mr. Traore Cheick Oumar

Abstract: Determining the budget revenue in big organizations, multinational company or government institutions, is a difficult tasks considering incertitude’s of some input items. In Sub-Saharan Africa countries, many governments still rely on manual, spreadsheet-based methods that are time- consuming, inconsistent, and error-prone to establish budget projections. This paper presents the design and implementation of a web-based budget revenue projection protocol system using artificial intelligence, developed for a general public organization responsible for national budget management. The goal of this research is to minimize the production time of the budget projection, minimize the projected data’ incertitude. The proposed protocol automates the projection of state budget revenues using statistical models (SARIMAX) and AI-driven analysis (Prophet), replacing previously manual forecasting methods.

DOI: https://doi.org/10.5281/zenodo.20307709

LPG Gas Detection System With Auto Cut-Off Regulator

Authors: Prof. M.D. Patil, Munjal Sumit Naganath, Shinde Atharv Maruti, Sayyad Azad Firoj

Abstract: The LPG Gas Detection System with Auto Cut-Off Regulator using Servo Motor is designed to improve safety by detecting LPG gas leakage and taking automatic preventive action. LPG is widely used in homes, hotels, restaurants, laboratories, and industries because of its efficiency and convenience. However, leakage of LPG gas can lead to dangerous situations such as fire accidents, explosions, suffocation, and property damage. To overcome these problems, the proposed system uses an MQ-6 gas sensor to continuously monitor the surrounding environment for LPG leakage. The sensor sends signals to the Arduino UNO microcontroller whenever the gas concentration exceeds the predefined safety limit. The microcontroller then activates a buzzer and warning indicator to alert nearby users about the leakage condition. This quick alert system helps people take immediate safety precautions and avoid hazardous situations. The system operates automatically with fast response time and minimum human intervention. Therefore, it provides a reliable and effective solution for improving LPG gas safety In addition to gas detection and warning generation, the proposed system also provides an automatic gas cut-off mechanism using a servo motor. The servo motor is connected to the LPG regulator knob and is controlled by the Arduino UNO microcontroller. When gas leakage is detected, the servo motor automatically rotates the regulator knob in the OFF direction to stop the gas supply and prevent further leakage. The system also uses relay modules to disconnect electrical appliances and activate an exhaust fan to remove leaked gas from the surrounding area, reducing the chances of fire and explosion. The proposed system is cost-effective, compact, and easy to install in domestic as well as industrial environments.

DOI: http://doi.org/10.5281/zenodo.20309495

Bioethanol from Agricultural Residues: Feedstock Characteristics, Conversion Pathways, And Engineering Challenges

Authors: Aditya Choukiker, Om Prakash Sondhiya

Abstract: Bioethanol remains one of the most important renewable liquid fuels because it can be blended with gasoline, distributed through existing fuel systems, and produced from a broad range of biological feedstocks. While first-generation ethanol relies on sugar- and starch-rich crops, increasing interest has shifted toward agricultural residues such as groundnut shell, sugarcane bagasse, rice straw, and corn stover. These residues are attractive because they are abundant, inexpensive, and do not directly compete with food use. Their conversion is nevertheless technically demanding because lignocellulosic materials contain cellulose and hemicellulose embedded within a lignin-rich matrix that resists hydrolysis. This paper presents a research-style review of residue-based bioethanol production with emphasis on feedstock structure, pretreatment methods, hydrolysis and fermentation pathways, product recovery, and practical engineering challenges. Groundnut shell is examined as a representative residue because it is readily available in many agrarian regions yet comparatively underused as an energy resource. The paper synthesizes published engineering and bioenergy literature into a coherent overview, compares selected residues on the basis of composition and process suitability, and discusses major barriers including pretreatment severity, enzyme cost, inhibitor formation, feedstock variability, and scale-up complexity. The review concludes that residue- derived bioethanol is technically feasible and environmentally relevant, but successful deployment depends on better feedstock logistics, process integration, and biorefinery strategies that improve carbon efficiency and reduce conversion cost

DOI: http://doi.org/10.5281/zenodo.20320778

Medical Visual Question Answering

Authors: Latheesha P, Meenakshi P, Nikitha D, Uma Maheswari T, Dr. Sundara Rajulu Navaneetha Krishnan

Abstract: A Transformer-Based Multimodal Framework for Medical Visual Question Answering offers significant convergence of computer vision and natural language processing for automatic medical image understanding with text-based inquiries. Common shortcomings of previous approaches are the requirements of interpretable outputs, small annotated datasets, and domain-specific reasoning under restriction. In this paper, we propose a transformer-based Med-VQA framework that is optimized on the Med-VQA-RAD dataset and whose structure is based on Salesforce-VQA-Base. Our approach takes advantage of the fusion of multimodal features-textual and visual data in enhancing response dependability and accuracy. We assess the performance using standard metrics like accuracy, BLEU, and medical-focused evaluation measures that demonstrate gains over baseline models. The results indicate that the proposed architecture enhances the diagnostic question-answering capability and offers understandable information to clinicians. The current work will lead the path for future research in scalable and explainable Med-VQA systems and will advance the development of AI-assisted tools for clinical decision-making.

Influence Of Personality Traits On Online Consumer Buying Decisions For Mobile Phones: Evidence From A Tier-2 Indian City.

Authors: Muhammed Muntaqheem G, Dr. J K Raju

Abstract: The exponential growth of e-commerce has fundamentally transformed consumer purchasing Behaviour, particularly in the mobile phone segment. While technological and economic factors have been widely studied, limited empirical research has examined the role of individual personality traits in shaping online purchase decisions, particularly in emerging urban markets. This study investigates the influence of Big Five personality traits on online consumer buying decisions for mobile phones, using primary data collected from 200 respondents in Davangere city, Karnataka. A structured questionnaire was administered, and data were analysed using Structural Equation Modelling (SEM) with Smart PLS. The results reveal that Openness (β = 0.32), Conscientiousness (β = 0.29), and Extraversion (β = 0.23) significantly and positively influence online buying decisions. Conversely, Neuroticism exhibits a negative but statistically insignificant relationship. The structural model explains 61% of the variance in online buying decisions.

DOI: https://doi.org/10.5281/zenodo.20323334

Hardware Root of Trust Based Secure Boot System for Embedded IoT Devices

Authors: Dr .D. Kumutha, Punith Raj R, Chinmayi TS, P. Sai Sushant, Ritesh Majjagi

Abstract: More and more IoT devices are turning up in factories, hospitals, homes and important systems. This raises serious security concerns for the hardware inside them. Things like firmware changes or malware can hit right at startup and that causes real trouble. Traditional software protections do not cover the early startup stage all that well. The idea here is to use a hardware root of trust to build a safer boot process for these embedded systems. Trust starts from something that cannot be changed in hardware and then moves up to the main firmware. It seems this chain helps prevent unauthorised access and rollback issues before they get underway. Cryptographic checks using SHA-256 and ECDSA help verify that the code is intact and comes from the right source. They built it around STM32 microcontrollers with a fairly light bootloader that fits limited hardware. I think this keeps things practical for smaller devices. Tests showed better resistance to tampering and more reliable checks overall. Protection against attacks improved in the results but there could be other angles to consider in real deployments.

Ai-Powered Corporate BI Sales Forecasting Dashboard Using XGBoost

Authors: Mohana Dharsan, Momin Ameer Basha, Nallagatla Venkata Ramanaiah, Mr. K. Senthilkumararaja

Abstract: This paper presents a production-grade AI-powered multi-item sales forecasting framework designed for corporate business intelligence (BI) environments. The system employs the Extreme Gradient Boosting (XGBoost) ensemble algorithm to generate accurate daily sales predictions across ten distinct product lines. A two-year historical dataset (2024–2025) comprising 7,300 records is processed through a temporal feature engineering pipeline that constructs eight predictive features: two lag variables (lag_1, lag_7), two rolling averages (7-day, 14-day), day-of-week, week-of-year, month, and a binary weekend indicator. Two complementary Streamlit-based interactive dashboards are implemented: (i) a Corporate BI Dashboard delivering a full-year 2026 annual forecast with five KPI metrics, trend visualization, monthly distribution charts, and CSV export; and (ii) an Advanced Forecasting Dashboard offering configurable 7–60 day horizons, model MAE transparency, actual-vs-predicted validation charts, and 95% confidence interval bands. A vectorized O(1)-append rolling-buffer forecasting loop enables 365-day prediction in under two seconds on commodity hardware. Empirical evaluation demonstrates distinct item-level demand patterns: item_1 exhibits a sharp January peak with −25.22% year-on-year decline, while item_2 shows stable +1.19% growth with an October demand surge. These insights enable data-driven inventory pre-positioning, promotional timing, and safety-stock calibration across enterprise planning horizons.

Footstep Power Generation For Smart Street Lighting And Charging

Authors: K H Suhas, Raghunandan V, Rajath S A, Srinivas H S, Dr. Sahana Raj B S

Abstract: The increasing demand for renewable and sustainable energy sources has encouraged the development of innovative energy harvesting technologies. Footstep energy harvesting is a promising method that converts mechanical energy generated by human walking into electrical energy. This work presents the design and implementation of a piezoelectric-based footstep power generation system capable of producing electrical energy from pedestrian movement. In the proposed system, multiple piezoelectric sensors are installed beneath a footstep platform where they experience mechanical stress whenever a person steps on the surface. Due to the piezoelectric effect, the applied pressure produces electrical voltage that can be collected and processed. The generated electrical output is passed through a rectifier and voltage regulation circuit to convert it into stable DC power suitable for storage. The conditioned energy is stored in a rechargeable battery and later utilized for practical applications. The stored power is used for automated street lighting and an RFID-based charging station that allows authorized users to charge small electronic devices.

DOI: http://doi.org/10.5281/zenodo.20324786

Assessing the Importance of Product Quality in Small and Medium Foundries of India

Authors: Mahantesh M Ganganallimath, Dr. K. Vizayakumar, Dr. Umesh M. Bhushi

Abstract: By providing crucial components to sectors like heavy engineering, construction, and automotive, the foundry industry contributes significantly to India's manufacturing economy. Although they make up a sizable share of this industry, small and medium-sized foundries (SMEs) confront formidable obstacles in terms of cost, quality, and worldwide competitiveness. This study evaluates the significance of product quality in Indian SME foundries, concentrating on its effects on sustainability, customer satisfaction, cost effectiveness, and productivity. The study emphasizes the need of systematic quality management methods by highlighting important quality criteria, obstacles, and options for development. The results show that product quality is a strategic component for long-term growth and competitiveness in addition to being a technical necessity. In order to assess overall quality performance, the study takes a quantitative and analytical approach, utilizing statistical quality tools, process capability analysis, and a Composite Quality Performance Index (CQPI) model. The results show that although many SME foundries attain acceptable product quality through rework and inspection, process quality is still relatively low because of excessive variability, insufficient control systems, antiquated machinery, and a lack of adoption of cutting-edge technology.

DOI: https://doi.org/10.5281/zenodo.20325463

The AI Music Mood Classifier Is Used To Classify Music Mood

Authors: Yashveer Singh, Vipin Dhiman, Shilpy Sharma

Abstract: The AI Mood-Based Music Classifier is a flexible microservice system that smartly sorts music into moods like happy, sad, angry, relaxed, romantic, energetic, chill, or focus by blending audio emotions from Librosa analysis, lyrics sentiment, and real-life context like weather or your activity—all powered by a Python FastAPI backend with NumPy crunching numbers, BullMQ queues on Redis for smooth async processing, and AWS S3 for storing admin-uploaded tracks that get transcribed and saved as metadata. Gemini LLM recommendations and live weather data are incorporated into the sleek Next.js frontend, which is built with TypeScript, Prisma ORM, and ShadCN to produce personalized playlists that can be played or adjusted on the fly.

DOI: http://doi.org/10.5281/zenodo.20338714

AI Symptom Assistant Checker: A Machine Learning–Driven Framework For Preliminary Healthcare Assessment

Authors: Hemlata, Ashish Prajapati, Satyam Shrivastava

Abstract: Studies have demonstrated the significance of early death cause interpretation for timely intervention in healthcare globally. The timely implementation of positive corrective measures for healthcare deaths has been correlated with healthcare studies. [1], [7]. Regrettably, the majority of healthcare professionals in resource-poor and neglected areas lack sufficient medical knowledge. Healthcare symptoms are taken at face value and delayed due to unrestricted online clues [2]. Artificial intelligence, or more specifically, predictive machine learning-based decision support systems, have made significant progress in recent years, and that's a positive thing. AI Symptom Assistant Checker is a web application that uses research and machine learning to make initial predictions about a disease by simply inputting its symptoms. It is like any other enabled healthcare AI systems, though in this specific instance we developed a responsive web application, coupled with a flask python web framework and a supervised machine learning model which are Decision Tree, Random Forest, Naive Bayes, Support Vector Machine and Neural Networks which have been used in other research to predict and classify for healthcare. [4], [5], [6]. Well-defined medical datasets are typically used to train and evaluate predictive healthcare models. The standard model for approaches in biomedical data analytics and clinical model development can be found here.

DOI: http://doi.org/10.5281/zenodo.20338735

A Specialist In AI-based Appointment Scheduling

Authors: Vishu, Vishal Arya, Rajesh Saxena

Abstract: The demand for appointment scheduling solutions that are smart, efficient, and scalable is rising in service-based industries such as healthcare, education, and professional services due to the rapid growth of digital technologies. In the past, appointment scheduling systems relied solely on manual coordination or fixed form-based scheduling solutions. Some of the challenges connected with traditional appointment scheduling solutions include conflict scheduling, late confirmations, poor resource utilization, a heavy administrative burden, and issues with user experience, especially within industries where growing demands for services have presented a challenge to efficient operations and service delivery. Automated decision-making by AI-based intelligent systems can improve appointment management systems through new scopes provided by recent developments in AI and web technologies. In this context, this research work proposes an AI-based appointment scheduling assistant built on the MERN technology stack, including MongoDB, Express.js, React.js, and Node.js, which is associated with natural language processing and rule-based AI technology. Through the use of natural language statements, the proposed work allows for appointment scheduling, creating a more user-friendly and interactive experience for clients. NLP techniques are employed by the project to extract important appointment details like date, time, and purpose, and then a rule-based AI engine checks appointment constraints and rules for compliance with pre-defined business rules. The implementation involves the use of JSON Web Tokens to ensure security for accessing the system and data integrity. All data related to appointments on the system is stored in MongoDB, a flexible and scalable storage system. The proposed system's experimental verification has shown improvements in scheduling accuracy, conflict resolution, human intervention, and system response time over the conventional scheduling method. The efficiency of a system can be greatly optimized by using NLP intelligence and expertise systems on a contemporary full-stack solution, as suggested by experimental results. The proposed work enables appointment scheduling in a cost-effective, intelligent, and efficient manner and sets the basis for performing improvisations in prediction, machine learning, and multiple language support.

DOI: http://doi.org/10.5281/zenodo.20338752

A Drug Discovery Platform That Is Powered By AI

Authors: Bhupendra Ram, Anurag Chandna, Sohan Lal

Abstract: Artificial Intelligence (AI) is the modern-day revolutionary force for drug discovery, offering a solution for the cost, time, and efficiency issues [1]. Unveiling a new drug through traditional pipelines takes over a decade and costs billions of dollars, and the high success rates in later stages have been failing [2]. The process of target identification, molecular design, and virtual screening is being transformed by AI-backed platforms and deep learning, graph neural networks (GNNs), and reinforcement learning (RL). The ability of algorithms to traverse large chemical spaces with greater precision and speed has been demonstrated by recent advances, such as AlphaFold in protein structure prediction and AI-aided molecule generation. The application of GANs and hybrid reinforcement learning methods to optimize molecules for both efficacy and safety is on the rise. In this paper, we present an overview of cutting-edge AI-enabled drug discovery platforms, highlight methodological advances, and propose a hybrid framework that integrates GNNs and generative models for efficient candidate optimization. Data privacy and replicability, as well as ethical and regulatory issues, are also discussed. Artificial intelligence drug discovery thus can lead to accelerated therapeutic development, cut costs, and enable personalized medicine advancements [3].

DOI: http://doi.org/10.5281/zenodo.20338788

AI-based Symptom Checker Assistant: A Data-driven Approach For Early Diagnosis And Digital Healthcare Support.

Authors: Dr. Yashveer Singh, Dr. Himanshu Tyagi, Dr.Ashish Saini

Abstract: DR. YASHVEER SINGH DR. HIMANSHU TYAGI Artificial Intelligence (AI) is transforming the healthcare where no other technology has had such a profound impact in treating patients, disease diagnosis, continuously monitoring patients in real time and dispensing medical aid. The AI-Based Symptom Checker Assistant is a new digital health tool that uses machine-learning algorithms to analyze user-provided symptoms and predict what diseases they may have. The purpose of this paper is to describe an AI symptom checker that was designed, implemented, and evaluated using a Random Forest classifier for disease prediction. Natural language input, structured data, automatic pre-processing, and an interactive user interface are some of the features of the system. The proposed model's high accuracy, user-friendly interpretation, and effective real-time response are demonstrated by the experiment results. The topic of AI symptom checkers in open primary health care is brought up, particularly in rural and low-resource regions.

DOI: http://doi.org/10.5281/zenodo.20338811

CNN And RNN Are Predicting A High-frequency Bit Coin Trend

Authors: Vivek Kumar, Vinod Rathi, Vineet Salar

Abstract: Bit coin is a type of digital currency that is used for online transactions. It is a digital currency that does not exist in hard currency form. Our focus is on the distinction between a decentralized currency and a centralized currency, which means that all virtual currency users can acquire services without the aid of a third party. Due to their severe price volatility, the use of these crypto currencies has an impact on international relations and trade. A reliable method for estimating this price is urgently necessary due to the rapid variations in the prices of crypto currencies. The level of one main or central control over them has been significantly affected by price control by a number of organizations, affecting relationships with other businesses and international trade. In addition, the constant oscillations suggest that a more precise method of estimating this price is urgently required. Thus, using deep learning techniques such as the recurrent neural network (RNN) and the long short-term memory (LSTM), gated recurrent unit (GRU), which are effective learning models for training data, we must design a method for the accurate prediction of by considering various factors such as market cap, maximum supply and, volume, circulating supply. Python is used to write the proposed method and it is tested on benchmark datasets. It can be inferred from the results that the proposed method is capable of making reliable predictions. For the past ten years, academics in various fields have used neural networks as one of the intelligent data mining tools. The importance of stock market data cannot be overstated in today's economy. Forecasting methodologies can be divided into two types: linear (AR, MA, ARIMA, ARMA) and nonlinear models (ARCH, GARCH, Neural Network). To anticipate a company's stock price based on past prices, we employed Autoregressive Integrated Moving Average (ARIMA), Recurrent Neural Network, Long Short-Term Memory (LSTM), and Gated Recurrent Unit Deep Learning Architectures (GRU).

DOI: http://doi.org/10.5281/zenodo.20338842

Using Model Explanation, Deep Learning Can Be Used To Detect And Classify Botnet Traffic

Authors: Mridula Singh, Shilpy Sharma, Sagar Chaudhary

Abstract: Malicious attempts known as distributed denial-of-service attacks because target services to be unavailable to legitimate users by sending many service requests that exceed the processing capacity of the services. Detection of botnet traffic is therefore critical to maintaining the availability and quality of the services while identification of the type of botnet attacks helps system administrators quickly figure out which part of the computer and network system are under attacks. The focus of existing research is on rule-based detection, which establishes rules in the network firewall to deny suspicious traffic that matches the rules. The emergence of machine learning and deep learning (ML/DL) has led to the development of preliminary works to learn botnet traffic behavior and perform detection. It is possible to enhance the performance of existing ML/DL models, but their decision-making and prediction are not transparent, which makes it hard for users to interpret and trust the results. In this work, we develop a novel deep learning model for botnet detection and classification, which has the ability to explain the model's decisions. Convolutional neural networks' latent representation of traffic feature allows us to detect if a traffic record is generated by a bot and then identify the type of bot responsible for the record. The predictions of the developed deep learning model are interpreted using an existing explainable framework. Extensive experiments are conducted with both realistic network traffic and synthetic traffic generated by the IXIA Breaking Point System. Various performance metrics are used to compare the developed model with existing models. The test results indicate that the developed model surpasses the current machine learning classifiers by up to 15% on all performance metrics, while providing a straightforward explanation of the model decision.

DOI: http://doi.org/10.5281/zenodo.20338856

An Explainable Product Recommendation System Using Artificial Intelligence And Machine Learning

Authors: Dr Chunnulal, Dr Satendra kumar, Dr Raj kumar

Abstract: Modern digital applications have incorporated recommendation systems, which enable platforms to provide users with personalized and relevant content. The aim of these systems is to predict user preferences by identifying patterns in large datasets, which will improve user experience and reduce information overload. E-commerce, movie and music streaming services, digital libraries, social networks, and online news platforms are among the many domains where they have been widely adopted. Prominent examples include Amazon recommending products based on previous purchases, Netflix suggesting movies based on viewing history, and music platforms like Spotify and Deezer generating customized playlists. A recommendation system generally gathers interaction data like ratings, browsing history, and time spent on content, then uses computational techniques to predict what users might want to choose next. The simplicity and strong predictive capabilities of collaborative filtering has made it one of the most widely used methods among various approaches. The assumption of collaborative filtering is that users with similar preferences in the past will have similar interests in the future. The filtering approach can be categorized as user-based and item-based. User-based filtering is used to identify users who have similar rating patterns, while item-based filtering is used to find similarities between items by analyzing how users interact with them. Python and collaborative filtering techniques are used to develop a movie recommendation system in this research. The dataset employed contains key attributes, including user IDs, movie ratings, item identifiers, and time spent on each item. Estimating how likely a user is to enjoy a particular movie is part of the recommendation process by analyzing similar users' preferences. After generating predicted ratings, the system ranks the movies and suggests those with the highest predicted scores. To create a functional recommendation system that can deliver personalized movie suggestions, collaborative filtering can be effectively implemented, as demonstrated in this study. Correlations and similarity measurements are used by the system to analyze user behavior patterns and provide relevant recommendations that enhance user engagement. Today's digital landscape is marked by the importance of recommendation systems, as shown by the results.

DOI: http://doi.org/10.5281/zenodo.20338889

Using Machine Learning To Detect Fake News

Authors: Reshoo Devi, Ashish Kumar, Baiju Kumar Yadav

Abstract: Fake news detection has become an urgent priority due to the widespread misinformation on digital platforms. A machine learning-based system is proposed in this research to classify news articles as real or fake using Natural Language Processing (NLP). The study utilizes TF-IDF vectorization and supervised learning algorithms like Logistic Regression, Naive Bayes, Random Forest, and Decision Tree to identify the most effective model for journalists, fact-checkers, and social media platforms. Logistic regression was discovered to be the most accurate model with 92% accuracy, demonstrating the power of machine learning in combating digital misinformation.

DOI: http://doi.org/10.5281/zenodo.20338924

Methods, Challenges, And Future Directions For Detecting Fake Reviews

Authors: Dr Raj Kumar, Viklpa, Rimmy

Abstract: The influence of online consumer reviews on purchasing behavior, market visibility, and perceived credibility of digital platforms is significant. Consumers' evaluation of product quality, service reliability, and seller trustworthiness is heavily influenced by review ratings and textual opinions, as demonstrated by empirical studies. Small changes in aggregated ratings can have a significant impact on sales volume, search rankings, and long-term brand reputation. Online reviews have become high-value informational assets due to this dependency, which makes review systems attractive targets for manipulation. Economic and social consequences have been significant due to the increasing prevalence of fake and deceptive reviews. Unfair competitive advantages are gained by businesses who engage in review manipulation, while honest sellers suffer revenue loss despite offering comparable or superior quality. Consumers who are exposed to deceptive reviews face an increased risk of poor purchasing decisions, wasted expenditure, and reduced confidence in online marketplaces. Repeated exposure to fraudulent content at a large scale can erode platform credibility, weaken user engagement, and undermine the integrity of digital ecosystems constructed on user-generated content. In modern review environments, manual moderation mechanisms, such as user reporting and expert inspection, are not effective. The processing of millions of reviews daily by large platforms makes human-driven verification costly, inconsistent, and slow. To maintain trust, fairness, and scalability, automated fake review detection has become a crucial requirement. Despite the challenges, designing effective detection systems remains a challenge. To resemble genuine user opinions, fake reviews are often crafted in a way that is linguistically fluent, sentimentally plausible, and strategically crafted. To evade detection, deceptive reviews take advantage of subjectivity, context dependence, and social norms, unlike traditional spam.

DOI: http://doi.org/10.5281/zenodo.20338959

Human Pose Estimation Using Deep Neural Networks

Authors: Anurag Chandana, Bhupendra Ram, Mukesh Tiwari

Abstract: Every tracking mechanism requires object detection where object tracking is the process in which locating an object or multiple objects is done using either the static or dynamic camera. It is important and challenging to detect and track objects in real-time. The recent focus of computer vision research has been on detecting and tracking multiple objects in dynamic environments. The position of a person or object in an image or video can be inferred using pose estimation, a task in computer vision. Pose estimation is a problem that involves determining the position and orientation of a camera in relation to a particular person or object. Identifying, locating, and tracking a number of key points on a given object or person is the typical way to do this. Corners or other significant features can be significant for objects, while in humans, these key points represent major joints like an elbow or knee. Tracking these key points in images and videos is the objective of our machine learning models. CNN can be utilized to detect yoga postures and their probability.

DOI: http://doi.org/10.5281/zenodo.20338984

Developing A Multi-agent AI System That Uses AI To Analyze Web Synchronization

Authors: Gautam Tyagi, Nisha Sharma, Bhanu Partap

Abstract: For investors, analysts, auditors, legislators, and researchers, financial records including quarterly reports, yearly 10-K filings, and regulatory disclosures include a wealth of information. Dense textual sections, financial figures, legal disclaimers, footnotes, and forward-looking assessments are all included in these agreements, which can total hundreds of pages. Their intricacy makes hand analysis slow, inconsistent, and prone to biased interpretation. While natural language comprehension has been enhanced by recent developments in large language models (LLMs), these models lack source-grounded reasoning and display hallucinations when dealing with lengthy, unstructured financial data. The AI Based Web Synchronise Analyzer, a multi-agent system that combines Retrieval-Augmented Generation (RAG), LangChain components, LangGraph-orchestrated decision routing, vector embeddings, document graders, hallucination evaluators, and a Streamlit-driven interface, is presented in this study in order to overcome these constraints.

DOI: http://doi.org/10.5281/zenodo.20339011

A Multi-Agent Medical Report Summarization System Using NLP And Specialist LLM Agents

Authors: Dr.Vinod Kumar, Vineet Salar, Monti Saini

Abstract: Medical records contain many messy details, making it difficult for patients or general health workers to follow them. Regular NLP tools fail when they jump across fields because each part requires its own know-how, rather than a single model handling everything. To use multiple smart agents at once, use fast LLaMA technology that is sped up by Groq chips, is an idea. The focus of one is on heart issues, another is on lung function, and the third is on mental health – all working together. A team lead agent gathers their insights into a clear wrap-up, giving a comprehensive diagnosis that anyone can grasp once they're done. The setup is powered by Flask and allows users to send in files, check results from smart modules, or retrieve the condensed version later. Testing demonstrates that it is more accurate, generates fewer false details, and runs faster than basic one-model tools. Findings suggest that splitting tasks among focused agents leads to deeper, smarter handling of tricky health records [1], [2]. Remote care settings benefit greatly from this approach, which aids patients in comprehending information, sorting diagnostic steps, and streamlining clinic operations.

DOI: http://doi.org/10.5281/zenodo.20339099

Explainable Heart Disease Prediction System Using Ma-chine Learning

Authors: Riya Kapil, Riyanshu Saini, Ashish Srivastava

Abstract: The majority of deaths worldwide are caused by heart disease. Doctors can use machine learning (ML) models to predict the risk of heart disease from routine exams and tests. The use of black-box ML models in healthcare is hindered by their lack of explanation for prediction. The proposed EHD-ML system combines effective ML models, such as gradient boosted trees and neural net-works, with techniques for explain ability that can be ap-plied to any model. SHAP, LIME, rule extraction, and counterfactual explanations are just a few of the things that are included. We cover dataset preparation, feature engineering, model training, interpretability pipelines, evaluation metrics like accuracy, AUC, F1, and calibra-tion, along with user-friendly explanations for clinicians, such as feature importance and patient-level explana-tions. We also outline the software and hardware design for deployment and suggest validation through retrospec-tive studies and prospective clinical trials. The key contri-butions include: (1) an end-to-end pipeline focused on explain ability for heart disease prediction, (2) a compar-ative analysis of interpretability methods and how accu-rately they reflect model predictions, and (3) user-centered explanation templates tailored for clinical use..

DOI: http://doi.org/10.5281/zenodo.20339115

Face Mask Detection Using Deep Learning

Authors: Rohan Chaudhary, Parul Tyagi, Neetu Maurya

Abstract: The COVID-19 pandemic emphasized the importance of face masks as an effective preventive measure to reduce the spread of airborne diseases. Ensuring proper mask usage in public areas through manual monitoring is inefficient, time-consuming, and prone to human error, especially in crowded environments. This paper proposes an automated face mask detection system based on deep learning and computer vision techniques to address these challenges. The system utilizes transfer learning with the MobileNetV2 architecture, selected for its lightweight design and high accuracy, making it suitable for real-time applications. OpenCV’s Haar Cascade classifier is employed for rapid face detection, after which the detected facial regions are classified into three categories: mask worn correctly, mask worn incorrectly, and no mask. Image preprocessing and data augmentation techniques are applied to enhance model generalization under varying lighting and pose conditions. The model is trained and evaluated using standard performance metrics, achieving high accuracy with minimal latency. The proposed system enables real-time monitoring and can be deployed in public spaces such as hospitals, transportation hubs, educational institutions, and workplaces. This approach reduces reliance on manual supervision and provides a scalable, cost-effective solution for automated health-safety monitoring.

DOI: http://doi.org/10.5281/zenodo.20339146

Explainable Covid 19 Severity Classification Using Deep Learning

Authors: Mayank Chauhan, Parul Tyagi, Jaishree Goyal

Abstract: Rapid identification of patients at risk of developing severe illness is necessary due to the major challenge presented by the COVID-19 pandemic to global healthcare systems. The purpose of this research is to propose a severity classification system that uses machine learning to predict the clinical outcome of COVID-19 cases at the time of diagnosis or early hospitalization. The study utilizes a structured dataset named “Covid Data.csv” comprising a wide range of demographic, clinical, and comorbidity-related features, such as age, sex, presence of pneumonia, diabetes, hypertension, obesity, ICU admission, and intubation status. Severity levels in terms of patient condition are reflected by the target attribute, 'CLASIFFICATION_FINAL'. In the data preprocessing phase, it was necessary to remove irrelevant records, add missing values, encode categorical features with Label Encoding, and normalize feature distributions by scaling continuous variables. Multi-class classification of severity categories was performed by a Random Forest classifier after preprocessing. Robust predictive performance with high accuracy, precision, and recall across all classes was demonstrated by the model. To assess misclassification patterns and validate reliability, a confusion matrix and classification report were created. Furthermore, the interpretation of the most significant predictors of severity was provided by feature importance analysis. StreamLight was used to integrate the trained model into a web-based user interface to ensure accessibility and usability. Healthcare professionals or users can use this application to input patient data through a form-like interface and receive predicted severity classification instantly, which can improve triage decisions and support early medical interventions.

DOI: http://doi.org/10.5281/zenodo.20339169

CodeGenie: A Multi-Agent Generative AI Framework for Explainable, Adaptive, and Scalable Automated Code Review

Authors: R. Hanush Singh, Dr. G. Maragatham

Abstract: Software code review is a labor-intensive process that directly influences the reliability, security, and maintainability of modern software systems. Conventional automated review tools, including linters, static analyzers, and monolithic large language model (LLM) assistants, suffer from limited contextual reasoning, narrow analytical scope, and a lack of explainability. This paper presents CodeGenie, a multi-agent generative artificial intelligence (AI) framework that decomposes the code review task into five specialized analytical roles: syntactic correctness, security, performance, stylistic consistency, and documentation quality. Each agent is implemented as a prompt-conditioned instance of a foundation generative model and operates in parallel on a shared code artifact. A weighted decision-fusion mechanism aggregates per-agent verdicts into a unified quality index, while an adaptive feedback loop personalizes review granularity based on developer history. The framework is evaluated on a curated benchmark of 480 source files spanning Python, JavaScript, and Java, drawn from open-source repositories and synthetic defect injections. Experimental results demonstrate that CodeGenie achieves a defect detection F1-score of 0.892, a 17.3% improvement over a single-model baseline, while reducing reviewer cognitive load by 41% in a controlled user study with 24 developers. The system contributes to United Nations Sustainable Development Goals 4, 8, and 9 by democratizing access to high-quality code review, improving developer productivity, and strengthening software infrastructure. We discuss the system architecture, mathematical formulation, empirical results, threats to validity, and future research directions toward self-improving agentic software engineering assistants.

AI-Based Waste Sorting System: A Research Review & Implementation Study

Authors: Dr Ashish Saini, Dr Amit Kumar, Dr Ankur Rana

Abstract: Rapid urbanization, population growth, and increased consumerism have led to a significant rise in municipal solid waste generation, creating serious challenges for effective waste management. Traditional waste segregation methods rely heavily on manual sorting, which is time-consuming, inconsistent, and exposes workers to hazardous materials. These limitations have motivated the exploration of automated solutions that can improve efficiency, accuracy, and safety in waste segregation processes. Recent advancements in deep learning, particularly Convolutional Neural Networks (CNNs), have demonstrated strong potential in computer vision–based waste classification tasks. CNNs are capable of automatically learningcomplex visual features such as shape, texture, and color, making them suitable for identifying different categories of waste materials. In this research, an AI-based waste sorting system was developed using a CNN model trained on publicly available waste image datasets similar to the TrashNet dataset. The system was designed to classify common waste categories, including plastic, paper, metal, glass, and organic waste. Experimental results show that the proposed model achieved an overall classification accuracy of approximately 90–92%, which is consistent with performance reported in recent literature. The findings indicate that a cost-effective and scalable AI-powered waste sorting system is feasible using existing deep learning techniques and affordable hardware components. This approach has the potential to reduce human involvement, improve recycling efficiency, and support sustainable waste management practices in urban environments.

DOI: http://doi.org/10.5281/zenodo.20339203

A Blockchain-Based Approach For Pharmaceutical Drug Authentication

Authors: P.Venkata Sai Shyam, D.Surya, N.Chintu, Assistant Professor Ms. Siva Selvi

Abstract: The pharmaceutical industry faces a critical challenge due to counterfeit and substandard medicines entering the supply chain. This project proposes a blockchain-based drug authentication framework that ensures transparency, traceability, and tamper-proof record management. By integrating Ethereum smart contracts and QR-code-based verification, manufacturers, distributors, pharmacies, and patients can verify the authenticity of medicines in real time.

AI-Based Career Guidance System For State And Central Government Examinations Using Multi-Agent Architecture

Authors: K. Kaethikeya Reddy, K. Naveen Kumar, E. Jithendra Reddy

Abstract: The rapid expansion of government employment opportunities in India has created a significant challenge for graduates and aspirants who lack structured access to personalized career guidance. This paper presents an AI-Based Career Guidance System specifically designed to assist students in identifying and preparing for state and central government examinations. The proposed system leverages a multi-agent artificial intelligence architecture inspired by the NexusAI framework, incorporating specialized agents for eligibility analysis, examination recommendation, syllabus guidance, and preparation strategy planning. The system integrates eligibility-based filtering, rule-based reasoning, and machine learning classification to deliver personalized examination recommendations aligned with the user's educational qualifications and interests. A web-based interface developed using HTML, CSS, JavaScript, and a Python backend with MySQL database management ensures accessibility and scalability. Experimental evaluation demonstrates that the proposed multi-agent approach enhances recommendation accuracy, user engagement, and preparation efficiency compared to conventional manual or single-agent advisory systems. The system addresses critical gaps in awareness, eligibility matching, and structured preparation planning for government examination aspirants.

Comparative Evaluation of Conventional and Sustainable Soil Stabilization Techniques for Ravine Slopes in the Bundelkhand Region, India

Authors: Vaishali Singh, Anupam Kumar Gautam

Abstract: Ravine slope instability is a persistent geotechnical and environmental challenge in the semi-arid Bundelkhand region of Uttar Pradesh, India. The problem is driven by shallow lateritic soils, steep slopes, intense monsoonal runoff, and progressive land degradation. This study presents a comprehensive comparative evaluation of conventional soil stabilization techniques—lime and fly-ash treatment—and sustainable alternatives, including vetiver grass bio-engineering and geocell reinforcement. Representative soil samples collected from Jhansi, Lalitpur, Banda, and Mahoba districts were subjected to laboratory testing to determine index properties, compaction behaviour, and unconfined compressive strength (UCS). Numerical slope stability analyses were conducted using limit equilibrium methods under both dry and saturated conditions. Results reveal that while lime stabilization produces the highest immediate strength gains, sustainable techniques achieve comparable improvements in slope stability with significantly enhanced erosion control and lower environmental impact. Factors of safety increased from 1.05–1.20 (untreated) to 1.50–1.70 (treated), depending on the method used. Sustainable methods demonstrated superior erosion reduction (60–80%) compared to chemical stabilization (40–60%).The study concludes that bio-engineering and geosynthetic-based approaches provide a balanced, environmentally sustainable, and economically viable solution for long-term ravine slope stabilization.

DOI: http://doi.org/10.5281/zenodo.20340981

AI Based Smart Agriculture Monitoring System

Authors: Mrs. R. Rajavaishnavi, V. Anand Kumar, U. Vamsi Reddy, T. Venkat Upendra Babu

Abstract: Smart Agriculture is transforming the conventional farming ecosystem through the integration of Artificial Intelligence (AI), Internet of Things (IoT), and cloud computing technologies. This paper presents the design and implementation of an AI-Based Smart Agriculture Monitoring System that continuously monitors critical agricultural parameters such as soil moisture, temperature, humidity, pH levels, and crop health using a distributed IoT sensor network. The collected sensor data is processed using advanced machine learning algorithms including Random Forest, LSTM networks, and MobileNetV2-based CNN models to deliver real-time insights and predictive analytics to farmers. The system computes unit-wise crop coverage weightage and allows flexibility through Auto Weightage, Equal Weightage, and Custom Weightage options. It also allows the inclusion of Bloom's Taxonomy cognitive levels to ensure balanced assessment of crop monitoring objectives. Finally, the generated monitoring report is formatted as per the required agricultural pattern and exported as a PDF file for field use.

Heart Safe: An Intelligent Wearable Cardiac Emergency Response System

Authors: Fasihuddin Siddiqi

Abstract: Heart Safe is an intelligent wearable healthcare concept designed to improve emergency response during cardiac emergencies through continuous physiological monitoring and automated communication technologies. The system was conceptualized with the objective of supporting individuals who may be at risk of heart attacks, cardiovascular abnormalities, or sudden medical emergencies. The proposed device continuously monitors important health parameters such as heart rate, blood pressure, pulse rate, and oxygen saturation levels. Whenever abnormal physiological conditions are detected, the system is designed to automatically trigger emergency alerts and transmit the user’s GPS location to predefined emergency contacts and healthcare responders. One of the key features of Heart Safe is its autonomous emergency response capability, which allows the system to operate even if the user becomes unconscious or unable to manually request assistance. The integration of wearable monitoring, GPS tracking, and communication technologies within a single platform aims to reduce delays in emergency response and improve patient safety. The project also explores the potential applications of Heart Safe in preventive healthcare, elderly care, remote patient monitoring, and connected healthcare systems. Furthermore, the report discusses the future scope, commercialization opportunities, technological advancements, and healthcare significance associated with wearable medical technologies. Overall, Heart Safe represents an innovative healthcare concept that demonstrates the growing potential of intelligent wearable systems in improving emergency healthcare support, preventive medicine, and patient-centered healthcare solutions.

DOI: https://doi.org/10.5281/zenodo.20343212

RFID Based Passport Authentication System

Authors: Prof. Krashna Rathi, Abhishek Jadhav, Shlok Shinde, Aniket Sarate

Abstract: The objective of this project is to develop an RFID-based passport authentication system using ESP32 technology to improve security and automate identity verification. The system uses RFID tags/cards as digital passports and verifies them through an RFID reader module (RC522). The ESP32 microcontroller acts as the main controller and processes the RFID data. If the scanned RFID card matches the stored authorized UID, access is granted; otherwise, access is denied. The system displays authentication status on an LCD display and indicates the result using LEDs and a buzzer. This system reduces manual verification, increases security, and provides a fast and reliable authentication mechanism.

Outland Survival: An Open World Survival Game

Authors: Kuna.Rakesh, K.Chandra Shekar Reddy, K.Thirupathiah, M.Vijay Baskar Reddy, Mrs.R. Mano Ranjani

Abstract: OUTLAND SURVIVAL is an open-world survival adventure set in a harsh and mysterious wilderness where nature, weather, hunger, and hidden dangers constantly challenge the player. After being stranded in an unexplored region known as the Outland, players must gather resources, craft tools, build shelters, hunt wildlife, and uncover the secrets buried within the land. The game combines realistic survival mechanics with exploration and freedom. Players can travel through forests, abandoned ruins, mountains, caves, and dangerous zones filled with hostile creatures and environmental threats. Dynamic day-night cycles and changing weather systems affect gameplay, forcing players to adapt their strategies to survive. As players progress, they can upgrade equipment, create settlements, tame animals, and discover ancient technologies hidden across the world. Every decision matter from managing food and health to ch oosing whether to fight, hide, or explore deeper into the unknown. With immersive environments, crafting systems, base building, and survival- focused gameplay, OUTLAND SURVIVAL delivers a challenging experience where survival is earned, not given. This project aims to demonstrate how advanced game technologies, including AI, physics simulation, and real-time environment processing, can be combined to create an interactive survival ecosystem. Performance evaluation and gameplay testing indicate that the system delivers stable frame rates, responsive controls, and immersive player experiences across different hardware configurations. The results highlight the effectiveness of the proposed architecture in developing scalable and engaging open-world survival games suitable for entertainment and research applications in interactive simulation environments.

An Efficient Email Spam Detection Framework Using TF-IDF Vectorization And Comparative Machine Learning Classifiers

Authors: Gandu Eshwar, Chirra Ram Gopal Rao, Thandu Venkat Sai

Abstract: Electronic mail is one of the most widely used forms of digital communication, yet it is increasingly compromised by the proliferation of unsolicited bulk messages, commonly referred to as spam. Spam email not only consumes bandwidth and storage but also exposes users to phishing, malware, and identity-theft risks. Conventional rule-based and blacklist-driven approaches struggle to keep pace with the rapidly evolving obfuscation strategies adopted by spammers. This paper presents an efficient and scalable email spam detection framework that combines Term Frequency–Inverse Document Frequency (TF-IDF) vectorization with a battery of supervised machine learning classifiers, including Support Vector Machine (SVM), Multinomial Naïve Bayes (MNB), K-Nearest Neighbors (KNN), Random Forest (RF), Extra Trees Classifier (ETC), and gradient boosted ensembles. Experiments performed on a publicly available labeled corpus of 5,572 messages demonstrate that the proposed TF-IDF + Linear SVM pipeline attains 99.9% accuracy on training data and 98.2% accuracy on unseen test data. Ensemble strategies based on soft voting and stacking achieve precision values as high as 1.0, eliminating false positives in the evaluated test partition. The reported findings establish the proposed framework as a lightweight, interpretable, and deployment-ready solution for real-world spam filtering systems.

Environmental And Economic Sustainability Assessment Of Plastic Waste-Based Pothole Repair Technology In Semi-Urban India

Authors: Umme Afsheen, Anupam Kumar Gautam

Abstract: Pothole formation severely impacts road infrastructure and road user safety in India, while the country grapples with massive non-degradable plastic waste generation. This study presents a comprehensive environmental and economic sustainability assessment of utilizing shredded plastic waste (LDPE, HDPE, and PET) in bituminous mixes for pothole repair using the dry process. The assessment is based on a real-world case study in Pratapgarh City and Chilbila, Uttar Pradesh. A functional unit of “1 km of pothole repair over a 5-year service life” was adopted for Life Cycle Assessment (LCA) following ISO 14040/14044 guidelines and Life Cycle Cost Analysis (LCCA). Results show that plastic-modified repairs divert 250–450 kg of plastic waste per km, reduce CO₂ emissions by 1.2–2.0 tonnes per km (45–55% reduction), and lower the overall environmental footprint by 30–40% compared to conventional methods. Economically, the approach achieves 18–22% savings in initial costs and nearly 50% reduction in lifecycle costs, with a Benefit-Cost Ratio (BCR) of 4.8 versus 1.4 for conventional repairs.The technology demonstrates strong alignment with circular economy principles, Swachh Bharat Mission, PMGSY, and Plastic Waste Management Rules. This study provides robust evidence for policymakers and municipal authorities to adopt plastic waste valorization as a scalable, sustainable solution for road maintenance in semi-urban India

Pov Display Drone

Authors: Dr. Jayashree Deka, Dr.Yogini Borole, Mr. Bhavesh Phegade, Mr. Mayur Hadpad, Mr. Shahejad Shaikh, Mr. Aditya Mogare

Abstract: This review paper explores the creation of an entertainment drone powered by a KK microcontroller, which includes a rotating LED display for dynamic visual effects and advertising purposes using an Esp32 microcontroller. The goal was to develop an affordable, adaptable, and programmable aerial platform that integrates stable flight control with captivating visual displays. The ESP32 is noted for its dual-core processor, built-in Wi-Fi and Bluetooth, and robust multitasking abilities, making it ideal for managing LED display tasks. KK is a common microcontroller used for drone integration. This paper covers the entire design process, from component selection to the integration of hardware and software. It also reviews existing point-of-view (POV) display drones and describes how the persistence of vision is employed to create midair visuals such as text, images, and animations. The drone can be controlled via remote, offering convenience and flexibility of the method The review addresses key challenges such as power management, display synchronization, flight stability, and the role of IoT in enhancing system performance. Practical applications, including light shows, advertising, and educational demonstrations are also discussed. Despite challenges like payload limitations and time sync, the review demonstrates that combining entertainment technology Embedded systems can offer economical and scalable solutions.

Real-Time Charge Analytics And Autonomous Fire Suppression For Electric Vehicles

Authors: Kudumula Kowshik Kumar Reddy, Kommineni Venkata Vamsi Krishna, Somarouthu Mohan Nagababu, Dr. K. Karthikeyan

Abstract: Challenges connected with temperature regulation in electric vehicle (EV) battery management systems (BMS) include overheating, delayed cold charging, energy depletion, efficiency degradation, cell imbalance, and durability concerns. These difficulties can be addressed by the measurement of voltage, current, and temperature of individual cells, real-time data processing via microcontrollers, state of charge (SoC) and state of health (SoH) estimations, thermal modelling, and defect detection. This paper presents a comprehensive BMS that monitors the SoC) and SoH of the battery pack, while integrating advanced thermal management and fire prevention features to mitigate potential hazards. The charging and discharging circumstances of the suggested approach are evaluated using the MATLAB Simulink tool. The outcomes of the hardware implementation and testing indicate that the proposed method enhances battery longevity and electric vehicle safety while also resolving critical issues related to charging efficiency and fire safety and temperature management in electric vehicles. This concept facilitates the development of safer, more intelligent, and more dependable battery systems for electric vehicles.

DOI: https://doi.org/10.5281/zenodo.20351850

Illegal Deforestation And Land-Use Change Detection From Satellite Imagery Using U-Net With CNN Backbone And Siamese CNN Architecture

Authors: Lalithavani K, Arulmozhi P, Vaidegi, Kiruthiga P, Nivetha S, Swetha E

Abstract: Background: Illegal deforestation and unlawful land-use change (LUC) represent two of the most severe anthropogenic threats to global biodiversity, carbon sequestration capacity, and ecosystem stability. Traditional field-based monitoring remains logistically impractical at continental scales. Objective: This study proposes a hybrid deep-learning framework that couples a U-Net segmentation architecture enhanced by a ResNet-50 convolutional neural network (CNN) backbone with a Siamese CNN designed explicitly for multi-temporal change detection, enabling automated, high-accuracy identification of illegal forest clearance and LUC events from multi-spectral satellite imagery. Methods: The pipeline is evaluated on three benchmark datasets—Amazon Deforestation Dataset (ADD), Sentinel-2 Global Land Cover (S2GLC), and the DESIS Hyperspectral Forest Dataset—covering more than 200,000 km² of tropical and sub-tropical biomes. Preprocessing integrates radiometric calibration, cloud masking, normalised difference vegetation index (NDVI) computation, and data augmentation to address class imbalance. Results: The combined framework achieves an overall accuracy of 96.8%, an Intersection-over-Union (IoU) of 0.923, and an F1-score of 0.947 on the held-out test partition, outperforming contemporary methods including DeepLab v3+, SegNet, and standalone Siamese networks. Conclusions: The proposed architecture demonstrates operational viability for near-real-time deforestation surveillance and may directly support regulatory agencies in enforcing environmental law under the Convention on Biological Diversity and national REDD+ frameworks.

DOI: https://doi.org/10.5281/zenodo.20352111

Design and Performance Analysis of an AI-Assisted Healthcare Web Platform for Remote Patient Support

Authors: Arkaprabho Saha

Abstract: AI-assisted digital healthcare systems are increasingly important for improving access, reducing administrative workload, and supporting early clinical decision-making. This paper presents the design and evaluation of CurePharm, a web-based healthcare platform that integrates symptom-based guidance, appointment management, and patient support functions. The proposed system combines a responsive web interface with backend logic for user interaction, data handling, and AI-assisted assistance. The objective is to study whether a lightweight engineering architecture can deliver reliable, scalable, and user-friendly healthcare support for common patient workflows. A prototype implementation was developed and evaluated using functional testing, usability feedback, and performance metrics. The results indicate that modular web architecture and AI-supported interaction can improve accessibility and streamline patient engagement. The paper concludes with design implications for future healthcare information systems and outlines opportunities for secure deployment and clinical integration.

An Integrated Charger For Wireless Power Transfer, Onboard Charger, And Auxiliary Power Module For Electric Vehicles

Authors: Dr.A.L.Renke, Ayush Katarkar, Sandeep Pawar, Shubham Patyekar, Sanket Suryawanshi

Abstract: Wireless power transfer (WPT) technology represents a transformative approach to electric vehicle charging, eliminating the requirement for direct physical cable connections while offering enhanced convenience, improved safety, and enabling dynamic charging capabilities that substantially extend effective driving range. This comprehensive review examines contemporary developments in wireless electric vehicle charging systems, synthesizing research across multiple technical domains including inductive power transfer principles, magnetic coupler design optimization, compensation network topologies, power electronic converter technologies, advanced control strategies, and infrastructure integration. Resonant inductive power transfer systems operating at standardized frequencies of 85 kHz have emerged as the most extensively developed and commercially viable approach, achieving demonstrated power transfer efficiencies exceeding 90% across practical air gaps of 150-200 mm [1]. The review systematically addresses critical technical challenges including misalignment tolerance between transmitter and receiver coils, electromagnetic field safety and regulatory compliance, and optimization of coil geometries to enhance coupling efficiency [2]. Recent innovations in dual and triple decoupled coil configurations maintain output voltage stability within 3% across ±150 mm lateral misalignment while achieving system efficiencies exceeding 94% [3]. Dynamic wireless charging systems enabling in-motion power transfer represent an emerging frontier, with advanced control strategies incorporating disturbance observers and adaptive frequency tracking maintaining power fluctuations within 0.2% despite vehicle motion [4]. Integration with renewable energy resources and smart grid infrastructure enables sustainable charging infrastructure with electricity cost reductions exceeding 36% compared to conventional grid-dependent systems [5]. Standardization efforts addressing interoperability between equipment from multiple manufacturers have achieved successful operation across diverse coil types with efficiency levels consistently exceeding 85% [6]. The comprehensive synthesis of contemporary research demonstrates that wireless electric vehicle charging technology has reached sufficient maturity for practical infrastructure deployment, with continued advancement focused on cost reduction, enhanced reliability, expanded interoperability standards, and seamless integration with renewable energy and intelligent transportation systems to accelerate widespread electric vehicle adoption and support global sustainable transportation objectives.

DOI: https://doi.org/10.5281/zenodo.20352437

Arduino Based Drowsinessand Fatigue Detection For Bikers Using Helmet

Authors: Dr.Punnya Priya F, Dinesh Karthikeyan S, Ruthra Kumar K, Rithu Krish S

Abstract: Improving vehicle safety is a key strategy used in addressing international and national road casualty reduction targets and inachieving safer road traffic comprises measures to help avoid acrash (crash avoidance) or reduce injury in the event of a crash(crash protection). Road traffic injuries are a major butneglected global public health problem, requiring concertedefforts for effective and sustainable prevention. Of all thesystems that people have to deal with on a daily basis, roadtransport is the most complex and the most dangerous.Worldwide, the number of people killed in road traffic crasheseach year is estimated at almost 1.2 million, while the numberinjured could be as high as 50 million – the combinedpopulation of five of the world’s large cities. What is worse,without increased efforts and new initiatives, the total number ofroad traffic deaths worldwide and injuries is forecast to rise bysome 65% between 2000 and 2020, and in low-income andmiddle-income countries, deaths are expected to increase by asmuch as 80%. This project deals with Drowsiness DetectionSystem.

DOI: https://doi.org/10.5281/zenodo.20354585

CaneCare: Real-Time Sugarcane Leaf Disease Detection Using Deep Learning — EfficientNet-B0, ResNet-50, And MobileNetV3 With Mobile Deployment

Authors: Roobannidhi K A, Revin Kumar R, Sakthi GaneshB

Abstract: Agriculture plays a vital role in the Indian economy, with sugarcane being one of the major commercial crops. Sugarcane is affected by diseases such as Mosaic, Red Rot, Rust, and Yellow Leaf Disease, which significantly reduce crop yield and quality. Early detection is essential for effective crop management, yet traditional manual inspection methods are time-consuming and inaccessible to rural farmers who lack expert guidance. This paper presents CaneCare, a deep learning-based system for automatic detection and classification of sugarcane leaf diseases from images. A curated dataset of 2,521 images across five classes—Healthy, Mosaic, RedRot, Rust, and Yellow—was used to train and evaluate three convolutional neural network architectures: EfficientNet-B0, ResNet-50, and MobileNetV3, all fine-tuned using transfer learning. ResNet-50 achieved the highest test accuracy of 94.71% (F1 = 0.947), while MobileNetV3 achieved 94.18% accuracy with only 1.5M parameters, making it the preferred choice for mobile deployment. EfficientNet-B0 provided balanced performance at 90.34%. A complete end-to-end deployment pipeline was implemented comprising a FastAPI backend hosted on the Render cloud platform and a React Native mobile application (CaneCare) built with Expo. Grad-CAM visualisations confirm that model predictions are grounded in biologically meaningful, disease-specific leaf features. The system provides real-time disease predictions with confidence scores and treatment recommendations, demonstrating a practical, accessible, and scalable pathway for AI-driven precision agriculture in resource-constrained environments.

FocusTrack- A Productivity and Focus Tracking App for Competitive Exam Students

Authors: Supesh Devche, Mrudul Dhakulkar, Bhumika Dewalkar, Samarth Deshmukh

Abstract: Long-term self-evaluation, disciplined study habits, and persistent focus are necessary for competitive exam preparation. Nonetheless, a lot of students find it difficult to stay consistent in their studies, control distractions, and accurately track their actual production levels. Current productivity apps are often made for broad work management; they seldom take into account the unique needs of students getting ready for competitive exams or offer insightful behavioural data about their study habits. This research introduces FocusTrack, a productivity and focus-tracking app made for competitive exam candidates. The system combines structured task management, Pomodoro-based session tracking, journaling, and visual analysis on a single platform. Users can create and organize study tasks, track their progress, log focused study sessions, and reflect on their learning through a digital journal. Additionally, features like experience points (XP), streak tracking, and visual productivity dashboards are included to promote ongoing engagement and motivation. Not like conventional assignment control equipment, FocusTrack emphasizes data-pushed productiveness evaluation by generating insights from look at period, undertaking crowning glory developments, and attention consistency styles. The software is carried out using a Flutter-based cell interface integrated through WebView with a subsequent.js internet utility, even as the backend infrastructure makes use of Firebase services which includes Authentication and Cloud Firestore for comfortable information garage and actual-time synchronization. A person-centric database architecture ensures privateness and relaxed get admission to to character productiveness information. The proposed device changed into evaluated through a pilot look at involving students getting ready for instructional and competitive examinations. Experimental observations imply enhancements in have a look at consistency, mission of completion conduct, and consciousness of private productiveness patterns amongst users. The outcomes suggest that FocusTrack can characteristic as a powerful virtual take a look at associate able to supporting disciplined learning and self-regulated take a look at practices.

DOI: https://doi.org/10.5281/zenodo.20355581

Fabrication & Modeling Of Multi-blades Areca Nut Dehusking Machine

Authors: Mr. Abhikumar C, MR. Abhishek gowda S, Mr. Abhilash M, Mr. Akash P, Ganapathy Bawge

Abstract: The multi-Blades areca nut dehusking machine is a semi-mechanized agricultural tool developed to reduce the labor and time required for dehusking Areca nuts and Coconuts. This machine operates using a DC motor and eliminates the need for automation, making it cost- effective and accessible for small scale farmers. It comprises a mechanical setup driven by a low- voltage DC motor, which powers dehusking sharp spike to strip the outer husk from Areca nuts and Coconuts. The manual feeding system ensures better control and safety. This report details the literature survey, objectives, methodology of the machine. Results indicate improved efficiency, reduced manual strain, and enhanced processing capacity compared to traditional manual methods.

Smart Safety Button For Accident Victims A One-Click Emergency Response System

Authors: Dr.Mahesh Nigade Atharv Gadache, Gaurav Babar, Tanvi Verma, Sanika Nangare

Abstract: Every year, millions of people around the world are caught in emergencies — road accidents, medical crises, or dangerous situations — and struggle to get help in time. A large number of these situations turn fatal not because help was unavailable, but because reaching that help took too long. Most of the current emergency tools, like calling 112 or pressing an SOS button on a phone, require the user to be calm enough to navigate their device. In a real crisis, that is often not possible. This paper presents the Smart Emergency Button, a mobile application designed to send instant emergency alerts with a single tap. The app is entirely software-based, meaning it works on any regular smartphone without requiring additional hardware. When the emergency button is pressed, the app automatically captures the user's live GPS location, identifies nearby hospitals, sends a pre-written emergency message via SMS and WhatsApp to saved emergency contacts, and displays directions to the nearest hospital on a map. The application was developed using standard web technologies and tested on multiple devices across different network conditions. Results show that the app can dispatch a complete emergency alert, including live location, in under 15 seconds on a standard mobile connection. Nine out of ten test users were able to use the app successfully without any prior instruction, confirming its ease of use. This paper describes the design, working, testing, and outcomes of this system, along with a discussion of its strengths, limitations, and potential for wider deployment.

Hand Gesture Robotic Arm

Authors: Aditiya Subba, Aryan Raj Shrestha, lidya Biswakarma

Abstract: The field of robots is now moving ahead rapidly. Robots are machines that perform physical tasks under the direction and control of humans. Many robots have been built to do dangerous jobs that humans can’t do on their own, and the robot arm is one of them. This research proposed a model and implementation of a robotic arm for controlled use with human hand and finger gestures. The movement mechanism of the robot arm is controlled by flexible sensors and fuzzy logic. Gyroscopes and human hand motions control the angle of the robotic arm. The input and output operations of the sensors, hand movement, and other controlling mechanisms are processed by means of fuzzy logic.

Thermodynamic And Kinetic Evaluation Of Crude Oil Bioremediation In Aqueous Systems Using Soursop Peel As A Sustainable Biocarrier

Authors: Dr. Malachy. O. Ugwuoke, Dr.Okoye Japheth. O, Dr. Agu Anthony

Abstract: This study investigates the thermodynamics and kinetics of crude oil bioremediation in aqueous systems using soursop peel as a natural biocarrier and Aspergillus niger as the degrading microorganism. Artificially contaminated water was treated over a 35-day period, and key parameters including total petroleum hydrocarbons (TPH), microbial count, and pH were monitored. Kinetic analysis revealed that the biodegradation process followed both first-order and pseudo first-order models. For the first-order model, a strong linear relationship was observed (R² ≈ 0.96) with a decay constant (k ≈ 0.006 day⁻¹), indicating that degradation rate depended on residual hydrocarbon concentration. Similarly, the pseudo first-order model showed excellent agreement (R² ≈ 0.98) with a decay constant (k ≈ 0.007 day⁻¹), confirming the influence of microbial activity and surface interactions. Thermodynamic evaluation using the Van’t Hoff plot (ln K vs 1/T) also exhibited good linearity (R² ≈ 0.96), validating the applicability of thermodynamic principles. The enthalpy change (∆H ≈ +28.5 kJ/mol) indicated that the process is endothermic, while the positive entropy change (∆S ≈ +0.095 kJ/mol·K) suggested increased randomness and enhanced microbial-substrate interaction. The Gibbs free energy change (∆G ranged from −12.6 to −9.8 kJ/mol) confirmed that the biodegradation process is spontaneous and thermodynamically feasible. Finally, the results demonstrate that soursop peel is an effective, low-cost biocarrier that significantly enhances microbial degradation of crude oil. The combined kinetic and thermodynamic findings confirm that the process is efficient, feasible, and temperature-dependent, making it suitable for sustainable remediation of hydrocarbon-contaminated water systems.

DOI: http://doi.org/10.5281/zenodo.20376931

Green Synthesis Of MnO₂ Nanoparticles Using Bunium Persicum Extract And Their Catalytic And Biological Applications

Authors: Jyotiba V. Pawar, Dhananjay V. Mane

Abstract: An eco-friendly and sustainable approach for the synthesis of manganese dioxide (MnO₂) nanoparticles was developed using Bunium persicum seed extract as a natural reducing and stabilizing agent. The green-synthesized MnO2 nanoparticles were characterized using UV–Vis spectroscopy, FT-IR, X-ray diffraction (XRD), scanning electron microscopy (SEM), and energy-dispersive X-ray analysis (EDX), confirming their crystalline nature, nanoscale dimensions, and high purity. The catalytic efficiency of the prepared MnO2 nanoparticles was evaluated in organic transformation reactions, where they exhibited excellent activity under mild conditions. In addition, preliminary biological studies demonstrated notable antimicrobial activity against selected bacterial strains. The results highlight the dual catalytic and biological potential of biosynthesized MnO2 nanoparticles and emphasize the advantages of plant-mediated nanomaterial synthesis in green chemistry and sustainable nanotechnology

DOI: http://doi.org/10.5281/zenodo.20377709

Correlation Analysis Between Token Price And Liquidity For Fraud Detection In DeFi Ecosystems

Authors: Dr. Pankaj Malik, Soham Bundela, Mohd. Ivaid, Mohnish Dhurve, Sharvi Saraswat

Abstract: The rapid expansion of Decentralized Finance (DeFi) has enabled open and permissionless token trading, but it has also led to a surge in fraudulent activities such as rug pulls, wash trading, and pump-and-dump schemes. This paper presents a novel fraud detection approach based on correlation analysis between token price and liquidity, leveraging the inherent relationship between these two market variables. In legitimate markets, price movements are typically supported by corresponding changes in liquidity, whereas fraudulent tokens often exhibit abnormal or decoupled behavior due to artificial price manipulation. To investigate this, we analyze time-series data of token price and liquidity across multiple decentralized exchanges and compute statistical correlation metrics alongside liquidity variation patterns. Experimental results show that legitimate tokens maintain strong positive correlations (r > 0.7) between price and liquidity, while fraudulent tokens exhibit weak or unstable correlations (r < 0.3), often accompanied by sudden liquidity withdrawals or artificial volume spikes. The proposed framework achieves high detection performance with an accuracy of 92.4%, precision of 90.1%, recall of 93.6%, and F1-score of 91.8%, demonstrating its effectiveness in identifying suspicious tokens at early stages. The findings confirm that deviations in price–liquidity correlation serve as a reliable and computationally efficient indicator for fraud detection in DeFi ecosystems. This approach can be integrated with existing blockchain analytics tools to enhance real-time monitoring and improve investor protection.

DOI: https://doi.org/10.5281/zenodo.20377805

Agentic AI

Authors: Shubham Jain, Akshay Chamoli

Abstract: The transition from conventional automation to agentic artificial intelligence represents a pivotal transformation in contemporary business operations, particularly within the B2B sales sector. Despite the widespread adoption of digital tools, traditional sales frameworks remain significantly constrained by operational inefficiencies, with sales development representatives (SDRs) dedicating up to 72% of their time to administrative and non-revenue-generating tasks while navigating fragmented and costly software ecosystems. These structural limitations not only impede productivity but also elevate operational costs and restrict scalability. Britha AI (also known as Gama AI) emerges as an autonomous, agentic platform designed to address these systemic challenges. Functioning as a digital employee, the platform leverages advanced agentic intelligence to independently manage and execute the end-to-end sales lifecycle, including intelligent lead generation, personalized multi-channel outreach, and the conversion of prospects into scheduled meetings. By consolidating previously disjointed tools and processes into a unified, agent-driven workflow, the platform fundamentally reconfigures the traditional sales pipeline, enhancing integration, efficiency, and overall performance. Adopting an empirical and data-driven approach, the analysis evaluates the platform’s architectural framework, unit economics, and go-to-market (GTM) strategy using real-world performance data. The findings indicate that the implementation of such an autonomous system can result in up to a 90% reduction in operational costs, primarily through the elimination of redundant tools and the minimization of manual intervention. Simultaneously, the platform demonstrates a tenfold increase in secured meetings, reflecting substantial improvements in outreach effectiveness and conversion rates.

Design A GPS-free Vehicle Tracking System Using GSM Module And Cell Tower Triangulation

Authors: A. Naresh, A. Sahithi, Sk. Naziya, D. Priyanka, B. Navya

Abstract: Modern intelligent transportation systems rely heavily on precise and efficient vehicle tracking to enable real-time monitoring, traffic management, and theft prevention. However, conventional GPS-based solutions often face limitations in areas with weak satellite coverage, high signal obstruction, or increased installation and maintenance costs. Existing alternatives, such as IoT-based hardware systems and Blockchain-enabled tracking, attempt to overcome these drawbacks but still suffer from higher latency, moderate channel utilization, communication overhead, and restricted location accuracy in dense urban settings. To address these challenges, this research proposed a GPS-Free Vehicle Tracking System that leverages GSM modules and cell tower triangulation for accurate and cost-effective vehicle positioning. The methodology incorporates triangulation algorithms that calculate vehicle coordinates using the signal strengths of multiple nearby towers, while an optimized channel allocation strategy ensures low collision rate, reduced transmission delay, and high request processing accuracy. Performance evaluation demonstrates that the proposed model consistently outperforms IoT Hardware and Blockchain & IoV approaches across all major metrics. Specifically, the system achieves an overall accuracy of 96.4%, precision of 95.9%, and F1-score of 96.2%, while maintaining the lowest transmission delay (1.3s) and collision rate (1.0). Furthermore, vehicle location tracking accuracy reaches 97.1%, validating the robustness of the triangulation approach in GPS-deficient environments. These results highlight the efficiency, scalability, and dependability of the GPS-Free system, positioning it as a practical and affordable alternative for real-time vehicle monitoring and fleet management applications, especially in regions with inconsistent or absent GPS coverage.

DOI: http://doi.org/10.5281/zenodo.20390055

Vision Transformer For Detection And Classification Of Microplastics In Water For Saving Aquatic Animals.

Authors: C.Sai Kalyani Deepthi, G.Dhana Lakshmi, SK. Najmabi, B.Lakshmi Mounika, K. Srividhya

Abstract: Quickly growing the microplastic contamination of the water body has emerged as a great ecological menace, compromising the marine biodiversity and water quality. Obvious limitations of traditionally used methods in detection, such as CNN-based models and other methods based on machine learning, include high computational cost, suboptimal accuracy in complex visual scenarios, sensitivity to the environment, and restrictions on the use in practice in real time. In order to overcome these drawbacks, this study suggests a Vision Transformer (ViT)-based architecture to effectively and precisely detect and classify microplastics on water images. The procedure starts with the further image enhancement based on Contrast-Limited Adaptive Histogram Equalization (CLAHE) that enhances the visibility of microplastics and reduces the noise. Images are split into patches and treated with transformer encoder layers with the help of multi-head self-attention to extract the global contextual information efficiently. A ViT-based decoder allows accurate classifying and segmenting of microplastics by type and size, which is trained on a hybrid loss based on cross-entropy and Dice losses to maximize pixel-wise accuracy. Experimental outcomes show that the proposed ViT model is better than traditional TinyML and CNN-based models, its detection accuracy is over 97, feature extraction accuracy is over 95, and its precision is 96.5, and the encoding time is cut by about 30. The model has strong generalization capabilities on a wide range of aquatic data with uniform training-validation results, which make it applicable to environmental monitoring. Overall, this ViT-based solution is scalable, computationally efficient, and highly accurate when it comes to assessing microplastic pollution, which enhances the conservation of the ecological environment, and offers a solution to real-time monitoring of the aquatic environment.

DOI: http://doi.org/10.5281/zenodo.20390106

A Novel Method For Air Quality-Driven Crop Prediction In Aeroponic Farming

Authors: M. Vasumathidevi, M. Lasya, B. Lakshmi Tirupathamma, Ch. Seema, K. Harshitha

Abstract: Traditional farming practices are not long-term viable due to the growing freshwater shortage and diminishing soil fertility, which pose serious threats to global food security. Because it eliminates soil-borne diseases, uses up to 95% less water, and supports sustainable development goals, aeroponic farming is a soilless cultivation method that emerges as an effective substitute. However, crop productivity in aeroponics is highly dependent on the quality of the surrounding air, in contrast to soil-based agriculture, where yield is determined by the fertility of the soil. In order to predict crop suitability in aeroponic systems, this paper proposes a novel methodology that combines the Air Quality Index (AQI) with real-time air pollutant concentrations of PM2.5, PM10, NO2, CO, and SO2. Based on scientific literature, AQI values and pollutant thresholds were mapped to crop tolerance levels to create a custom dataset. The suggested system uses a Random Forest classifier to suggest crops that are most suited to the local air conditions, calculates AQI using the US-EPA formula, and processes pollutant inputs. The model performs reliably, achieving 80% prediction accuracy while remaining robust to noisy data inputs. The study presents a scalable, flexible framework for crop selection that promotes sustainable agriculture and emphasizes the significance of air quality as a crucial component of soilless farming. In order to further improve reliability, future scope will involve adding more crops, incorporating environmental variables like temperature and humidity, and cross-referencing predictions with yield data collected at the field level.

DOI: http://doi.org/10.5281/zenodo.20390151

Save Vision: Smart Drive Monitor

Authors: Deeya Kharediya, Ayushi Gupta, Ankita Shrivastava, Manish Kr. Suman

Abstract: Driver fatigue, drowsiness, and cognitive distraction are among the most underreported yet dangerous causes of road accidents globally. Unlike impairment from alcohol or mechanical failure, these states are invisible to external observers and develop gradually, leaving little time for corrective action once a critical threshold is crossed. Despite advancements in vehicle safety engineering, a reliable, affordable, and non-intrusive system capable of detecting these states in real time remains largely inaccessible to everyday drivers — particularly in developing nations where road fatality rates continue to rise. Save Vision: Smart Drive Monitor is a camera-based, real-time driver monitoring system developed to bridge this gap. The system continuously captures the driver's facial region through a webcam and applies computer vision techniques to analyse four behaviorally significant risk indicators: prolonged eye closure as a marker of drowsiness, yawning as an early physiological signal of fatigue, head pose deviation as an indicator of distraction,and mobile phone usage as a documented cause of inattention. Detection is performed through facial landmark extraction, Eye Aspect Ratio (EAR) computation, and image-based classification signal of fatigue, head pose deviation as an indicator of distraction,and mobile phone usage as a documented cause of inattention. Detection is performed through facial landmark extraction, Eye Aspect Ratio (EAR) computation, and image-based classification, all processed entirely through software. Upon identifying a risk condition, the system activates a layered alert mechanism — combining an auditory alert through the system speaker and an on-screen visual warning — designed to capture the driver's attention proportionally to the severity of the detected state. The system accepts both live webcam input and pre-recorded video, making it flexible for real-world deployment as well as controlled testing environments. Save Vision is designed with accessibility and practicality at its core, targeting deployment without dependency on expensive hardware or cloud infrastructure. The system establishes a functional foundation for intelligent, preventive road safety — one that can be incrementally extended toward integration with advanced driver assistance systems and fleet-level monitoring platforms. By addressing driver risk at its earliest behavioral signs, Save Vision aims to contribute meaningfully to the reduction of road accidents caused by human inattention.

DOI: https://doi.org/10.5281/zenodo.20390165

Quantum Natural Gradient Based Collaborative Task and Systems Allocation Approach for Heterogeneous Multi-Unmanned Vehicles

Authors: T. Navya Deepthi, G. Krishnaveni, T. Sarika, M. Mounika, S. Tejaswini

Abstract: Task allocation is a critical component of various systems, requiring efficient and accurate assignment of tasks to resources. Existing models often struggle with balancing precision and recall, leading to suboptimal performance. The limitations of existing task allocation models, including low precision and recall, hinder the efficiency and effectiveness of task allocation processes. Current models often rely on simplistic approaches, failing to account for complex task dependencies and resource constraints. This leads to reduced accuracy and increased errors in task allocation. The Quantum Natural Gradient Based Collaborative Task and Systems Allocation Approach for Heterogeneous Multi-Unmanned Vehicles framework addresses these limitations by leveraging advanced algorithms and techniques to optimize task allocation. By integrating precision and recall metrics, Quantum Natural Gradient Based Collaborative Task and Systems Allocation Approach for Heterogeneous Multi-Unmanned Vehicles ensures a balanced approach to task allocation, minimizing errors and maximizing efficiency. Experimental results demonstrate the effectiveness of the Quantum Natural Gradient Based Collaborative Task And Systems Allocation Approach For Heterogeneous Multi-Unmanned Vehicles framework, achieving a precision of 96%, recall of 95%, and F1 score of 95.5%. The framework also exhibits a low task allocation time of 2.0 seconds and high pre-processing accuracy of 98%.

DOI: http://doi.org/10.5281/zenodo.20390184

Post-Quantum Blockchain Framework For Securing The Social Internet Of Things

Authors: V. Lakshman Narayana, K. Akhila, D. Mani, K. Rajeswari, B. Sai Vaishnavi

Abstract: The Social Internet of Things (SIOT) integrates social networking principles with IOT, enabling autonomous, collaborative interactions among smart devices. while security and privacy challenges due to Heterogeneity Of devices distributed communication. Traditional blockchain-based solutions provide decentralized trust and tamper-proof transaction recording, but they largely rely on classical cryptographic algorithms, which are vulnerable to quantum computing attacks. Additionally, existing models suffer from high key generation time, inefficient block creation, and low transaction validation accuracy, limitation their scalability and reliability in large IoT networks. To address these limitations Post- Quantum Blockchain Framework for Securing Social IoT (PQBCF-SSIoT) is proposed. This framework integrates post-quantum cryptographic algorithms with optimized blockchain protocols, ensuring quantum-resistant security, efficient key generation, and reliable transaction processing. PQBCF-SSIoT enhances block creation and validation accuracy while minimizing computational overhead, making it suitable for resource-constrained IoT devices. PQBCF-SSIoT achieves a block creation accuracy of 97.5% and a block validation accuracy of 98.2%, significantly outperforming SSIoT-GAN, which reaches 94.2% creation accuracy and 95.1% validation accuracy, and OAD2D-SIoT, which achieves 92.8% creation accuracy and 93.5% validation accuracy. The proposed model achieved better performance in providing security levels.

DOI: http://doi.org/10.5281/zenodo.20390235

Based On Quantum Cryptography, Blockchain, And Dcnn To Determine Emotions From Facial Expressions

Authors: V. Sujatha, B. Sri Sai Kavya Samhitha, B. Vyshnavi, M. Mallika, P. Teja

Abstract: Facial expression-based emotion recognition plays a crucial role in sectors such as healthcare, security, education, and human-computer interaction, yet achieving accuracy, privacy, and security remains challenging. This research proposes a secure and reliable emotion recognition framework by integrating Deep Convolutional Neural Networks (DCNN), blockchain technology, and Quantum Cryptographic Distribution (QCD). The framework employs a DCNN trained on diverse datasets to ensure resilience against variations in lighting, occlusion, and facial diversity. It captures facial images or video frames, extracts spatial and texture-based features, and utilizes these features for emotion classification. QCD provides post-quantum cryptographic security for data transmission, while anonymized results, timestamps, and hashed feature vectors are securely stored on a permissioned blockchain. Experimental evaluation on benchmark datasets (FER2013, CK+, and JAFFE) demonstrates superior performance compared to traditional CNN, VGG16, and ResNet50 models, achieving 97.8% accuracy, high precision (~0.98), recall (~0.98), and F1-score (0.98). The system is suitable for privacy-sensitive applications as it ensures tamper-proof, auditable, and real-time emotion recognition. By combining blockchain immutability, quantum-secured communication, and AI-driven classification, the proposed framework offers a secure, interpretable, and future-proof solution for emotion recognition in practical applications.

DOI: http://doi.org/10.5281/zenodo.20390297

Partial Replacement Of Coarse Aggregate With Coconut Shell And Cement With Sugarcane Bagasse Ash In M25 Grade Concrete

Authors: Aalok Kumar Yadav, Mr. Daljeet Pal Singh

Abstract: The increasing demand for construction materials has resulted in rapid depletion of natural resources and environmental degradation. This study investigates the use of Sugarcane Bagasse Ash (SCBA) as partial replacement of cement and Coconut Shell (CS) as partial replacement of coarse aggregate in M25 grade concrete. Different replacement levels of SCBA and coconut shell were used to study workability and compressive strength characteristics. Experimental results indicated that the use of 10% SCBA and 10–20% coconut shell provided satisfactory strength and durability properties while reducing environmental impact. The developed concrete can be considered sustainable, lightweight, and economical for future construction applications.

Adaptive Tree-Based Ensemble Framework For Real-Time Cyber Threat Classification (TBE-CTCF)

Authors: M. Francis, M. Shareena, K. Niharika, P. Nikhatjahan, Sk. Mehavish

Abstract: Traditional Cyber threat detection system depend on static,pre-trained models that fails to adapt changing patterns,leading to performance deterioration against zero-day threats. An adaptive real-time essemble framework (AREF) for cyber threat categorization is presented in this study to get over this restriction. It is intended to improve detection accuracy and flexibility by integrating dynamic models. Three machine learning classifiers are used by AREF to collaboratively process network traffic data in real time: XGBoost, LightGBM, and Random Forest. Different feature viewpoints are captured by each model, and their predictions are adaptively merged using a weighted stacking method that is adjusted by ongoing performance monitoring. Three models are used in this technique. Capturing nonlinear connections is the first step.in high-dimensional traffic characteristics while guaranteeing strong generalization against overfitting. By using leaf-wise growth with depth limitations and histogram-based gradient boosting, LightGBM speeds up real-time classification, allowing for quicker convergence and effective management of massive streaming data. Random Forest lowers variance and improves robustness to noisy and unbalanced datasets by introducing feature randomization and parallel decision aggregation.The framework may change in real time because to its adaptive ensemble technique, keeping its excellent accuracy even when network activity patterns change. According to experimental evaluation, AREF provides a scalable and explicable solution for real-time cyber threat detection and classification in dynamic environments, consistently outperforming static ensembles and individual base models in terms of F1-score, detection precision, and response latency.

DOI: http://doi.org/10.5281/zenodo.20391322

Cross-Modal Dynamic Hypergraph Attention Network For Early Detection Of Autism Spectrum Disorder

Authors: C. Sai Kalyani Deepthi, N. Jyoshnavi, N. Varshini, K. Harshitha, M. Ankitha

Abstract: Autism Spectrum Disorder (ASD) is a complex condition that affects social communication linked to repetitive behaviours. Early detection of ASD is crucial because timely treatment can significantly improve developmental outcomes. Traditional screening methods are often slow, subjective, and depends on single data sources like behavioural questionnaires or facial analysis. To solve these problems, a novel framework Cross-Modal Dynamic Hypergraph Attention Network (CDHAN) has been proposed for ASD identification. This framework combines facial image data with behavioural screening information to capture complex interactions between different data types. This model employs dynamic hypergraph structure to capture the higher-order interactions and uses attention mechanisms to focus on important features like eye-gaze direction, facial asymmetry, emotional expressiveness, response latency, and social interaction cues, resulting in reliable and easy-to-understand predictions. Extensive testing shows that CDHAN outperforms than previous models in accuracy of 97.21%, precision of 95.73%, recall of 96.38%, and F1 Score of 95.47%, specificity of 95.84% while keeping a low error rate of 4% and achieving high generalization across various datasets. By providing an automated, scalable, and clinically useful approach and allows for quicker and more reliable ASD screening and early therapies that can improve children's development.

DOI: http://doi.org/10.5281/zenodo.20391418

Hybrid Quantum-Classical Framework For Cyber Fraud Detection With Quantum Feature Selection Using Q-Defender Net

Authors: M. Satya Vijaya, J. Rinithya, T. Venkata Nandhini, K. Sharlee, B. Rushitha

Abstract: Cyber fraud is still a serious threat to data-driven infrastructures, e-commerce sites, and financial systems. It frequently evades detection models that use static rules or traditional machine learning. In order to improve detection accuracy and cost sensitivity, A method is present Q-Defender Net, a hybrid quantum-classical framework that blends ensemble classification and quantum feature selection. After preprocessing the data with normalization and class balancing, the system maps feature into Hilbert space using quantum kernel alignment, and then uses QAOA to identify the most informative features. The parallel classifiers XGBoost and Quantum SVM then process these features, and their outputs are combined using weighted voting.High-value fraud cases are given priority by a cost-conscious loss function, which enhances practical impact. According to experimental results, Q-Defender Net outperforms FD4QC and Hybrid ML in terms of error rate and convergence speed, achieving 99% accuracy, 98% precision, and a 98.5% F1-score. It is a potent remedy for contemporary cybersecurity issues due to its modular, scalable architecture and adversarial robustness.

DOI: http://doi.org/10.5281/zenodo.20391671

ResNet-50 Driven Deep Learning Framework For Diabetic Macular Edema Detection In Retinal Fundus Images

Authors: Y. Swathi, B. Haseena Parveen, M. Venkata Lakshmi, U. Harika, N. Sai Vybhavi

Abstract: Diabetic macular edema (DME) is an advanced stage of Diabetic Retinopathy (DR), caused, often resulting in vision impairment or that can lead to vision loss or blindness if not detected early. Timely and accurate detection of DME is critical for preventing vision loss and improving patient outcomes. In this study, a ResNet-50 based on deep learning framework is proposed for the automated detection and grading of DME from retinal fundus images. The model leverages the residual learning architecture of ResNet-50 to capture multi-level hierarchical features, enabling precise identification of pathological regions in the retina. The suggested model was thoroughly tested on several reference retinal datasets and contrasted with the most recent cutting-edge techniques. With impressive results, including 98% Accuracy, 97% Precision, 98% Recall, and 97.5% F1 Score, it outperformed current methods. Excellent discriminative ability between DME-positive and DME-negative cases is confirmed by the ROC curve analysis, and dependable and consistent predictions are highlighted by the confusion matrix. The robustness of the model is confirmed by the training and validation loss curves' steady convergence.

DOI: http://doi.org/10.5281/zenodo.20392019

Smart Helmet With Sensors For Accident Prevention

Authors: Mr.A.Prasad, Dr.Prasanna Murali Krishna, Dr.A.ranganayakulu, Yarramsetty Venkata Sivakrishna, Nune Yahosuva, Alisetty Sai Kumar

Abstract: The goal of this project is to develop and deploy an Internet of Things (IoT)-based intelligent system for environmental monitoring and accident reporting. The device combines accelerometer data over a particular level to detect car accidents, and data from a flame sensor to identify fires. Once an incident is identified, the system notifies an emergency response team via a GSM module and utilizes GPS to locate the exact location. urthermore, the system uses the appropriate sensors to monitor factors such as temperature, humidity, alcohol, smoke, and carbon monoxide levels. Upon receipt, the IoT platform has the option to evaluate and provide these data points in real-time. This integrated method improves safety by allowing for fast response to accidents and environmental monitoring; it might be used in smart transportation systems. Thingspeak, sensors, the internet of things, global positioning systems, data monitoring, the ESP8266, and danger alerts are some of the facets discussed.

DOI: https://doi.org/10.5281/zenodo.20392311

Design And Implementation Of RV321 Processor Core In VLSI

Authors: Dr.A.Ranganayakulu, Dr.D.Satyanarayana, Venna Kasieswari, Dudekula Rijvana, Duddela Ramya, Shaik Farida

Abstract: A single-cycle processor with support for 30 instructions based on the RV32I instruction set architecture (ISA) is presented in this research. Because it was necessary to fetch, decode, execute, and finish each instruction within one clock cycle, the single-cycle architecture was selected. When simplicity and power economy are more important than execution speed, this strategy works well for small-scale applications. Important parts of the architecture include a data memory for storing information, an arithmetic logic unit (ALU) for performing logical and mathematical operations, a register file for storing instructions, and instruction memory for storing data. In order to control the flow of data and make sure instructions are executed correctly, control signals are created depending on the opcode. The Verilog hardware description language (HDL) was used in the development of the CPU. As a platform for implementation, the Arty-S7-FPGA board was the goal of the design synthesis. Basic processes including data transmission, math computations, and control flow are effectively handled by the architecture. The test bench is built using verilog and all thirty instructions are tested in the Vivado software environment for operation verification. This research explores the RISC-V architecture from the bottom up and paves the way for improved implementations of pipelined processors in the future. Nouns: Vivado, Verilog, Single-Cycle Processor, RV32I Instruction Set Architecture.

DOI: https://doi.org/10.5281/zenodo.20392389

Smart Indoor Air Quality Monitoring System

Authors: Mr.M.Ramana Reddy, Mr.B.Ajantha Reddy, Sampeta Sai Kumar, Talluri Lingeswara Rao, Shaik Jani Basha, Kanigiri Venkateswara Reddy

Abstract: Temperature, humidity, PM 2.5, and formaldehyde are just a few of the air quality factors that may be measured in real-time by a low-cost system that utilizes LoRa technology. A star topology is used in the system's architecture, with sensor nodes linked to a routing node that is linked to a gateway node. The sensor nodes are built using inexpensive, commercially available sensors that are connected to a microcontroller. The sensor data is sent to the routing node, which is also built using a low-cost microcontroller (like Raspberry Pi), using a LoRa hat module SX1262. The gateway device stores, displays, and processes data using a time-series database and a server. A transmission range of around 3 km with no packet loss and 4 km with 20% packet loss may be accomplished using LoRa technology.

DOI: https://doi.org/10.5281/zenodo.20392482

Smart Vision System For Driver Alertness Detection

Authors: Dr.P.Prasanna Murali Krishna, Dr.D.Satyanarayana, Yammani Najarayudu, Shaik Mahaboob Sarif, Betham Prabhudas, Venna Lakshmireddy

Abstract: Fatigue and drowsiness are major contributors to automobile accidents, which kill over 1.3 million people annually. With the use of facial landmark detection, the Advanced Drowsiness Detection System is able to accomplish the aforementioned framework, which in turn reduces the total number of road accidents and the weariness that causes drivers to doze off. This technology uses a facial recognition algorithm to identify signs of sleepiness in a user's face. It finds the driver's face and follows its movements to calculate the Eye Aspect Ratio (EAR), a verified way to identify sleepiness. This technology utilizes Driving Behaviour Analysis (DBA) to identify when drivers are drowsy, which in turn decreases deaths and increases road safety. The goal of this technology is to make transportation safer and to prevent drivers from becoming sleepy while operating.

DOI: https://doi.org/10.5281/zenodo.20392582

Low Power RISC-V Embedded Processor Design Using HDL

Authors: Mrs.N.Swarupa Rani, Mr.N.B.Jilani, Ravuri Venkata Koteswararao, Shaik Irfan, Kancharla Chandu, Kuricheti Pavan Sathwik

Abstract: We are entering a new age, and with it comes a surge in interest in generative AI and machine learning throughout the world. As computing talents start to reveal their true colors, there is a pressing demand for efficient and fast microprocessors. The RISC design approach, which stresses a smaller and simpler set of instructions, each executed in a constant amount of time, is widely used by current processors to achieve improved efficiency. A five-stage pipelined general-purpose microprocessor architecture with a clock period of 0.59 nanoseconds has been the target of this study, which translates to optimization approaches being implemented in Verilog code. In addition, by switching to True Single- Phase Clocking (TSPC) flip-flops, we may cut power usage by 20% and cut the number of transistors used by half. The use of Verilog HDL on ModelSim and Xilinx to achieve an n-stage pipelined architecture for a general-purpose microprocessor is shown in this work as an example of RISC-V embedded processor design.

DOI: https://doi.org/10.5281/zenodo.20392646

High Speed 32-Bit Vedic Multiplier Using Pipelined Architecture

Authors: Mr.N.B.Jilani, Dr.P.Prasanna Murali Krishna, Perumalla Geeta Bharathi, Venna Leelavathi, Porumamilla Dharshini, Sure Anusha

Abstract: Based on the historical practice of Vedic multiplication, this proposed study details a large-number variant of the Vedic multiplier. Since it improves over the alternatives in terms of speed, area, hardware complexity, power consumption, and scalability, a Vedic multiplier is the way to go. While Xilinx is used for circuit design, the Modelsim tool is used for implementation using Verilog HDL. The Carry Save Adder, the Carry Lookahead Adder, and the Ripple Carry Adder are three 32-bit Vedic multiplier designs that are analyzed in this paper. They use different adder architectures. Around 0.082W was the power consumption of all three designs. With a latency of 26.466 ns and almost similar power usage, the 32-bit CLA-based Vedic multiplier was the quickest. The RCA-based multiplier outperformed all other Vedic multipliers in terms of area usage. Search Terms—Circuit Design, Power Consumption, Vedic Mathematics, Parallel Processing, and High-Speed Multiplication [4]. The procedure for practicing vedic multiplication is shown in Fig. 1. By dividing the multiplication operation into smaller, more manageable parts, this method offers a new solution to the problem. When compared to older multiplier designs, this may provide better results, particularly when taking into account the power and performance limitations of contemporary DSP systems.

DOI: https://doi.org/10.5281/zenodo.20392712

TrustEdu: A Blockchain-Based Framework for Secure Educational Document Verification

Authors: Soham Baban Kale, Nikhil Anil Khose, Vaibhav Uttam Kolekar, Vedant Vaibhav Kondejkar, Prof. Amol Jagtap

Abstract: Document verification has always been a manual, time-consuming task for institutions and companies alike. According to data released by India's Ministry of Education, close to one million students graduate every year and move on to either higher studies or employment. Each of them carries a set of academic records — mark sheets, certificates, diplomas — that need to be verified by receiving institutions or employers. The problem is that current verification systems are centralized, which opens them up to tampering, SQL injection, collusion attacks, and straightforward document forgery. There is no reliable mechanism in place that gives a third party instant confidence in a document's authenticity. This paper presents TrustEdu, a web-based digital document locker built on top of a custom blockchain. The system lets students upload their academic documents, get them verified by their institution, and then share a QR code or unique instrument ID with any third party — eliminating the need to carry physical documents altogether. Once a document clears institutional verification, its hash is stored on the blockchain, making any subsequent tampering immediately detectable. Smart contracts automate the verification logic, and the QR code serves as a tamper-evident link between the paper record and its blockchain entry. Results show that the system significantly reduces verification time while improving the integrity and trustworthiness of shared academic credentials.

DOI: https://doi.org/10.5281/zenodo.20392948

Trophic Transfer of Industrial Heavy Metals from Polluted Irrigation Water to Food Chain: A Public Health Risk Assessment in Unnao, India

Authors: Dr Amit Kumar Awasthi

Abstract: Unnao district in Uttar Pradesh, India, located along the Ganga Plain, has emerged as a critical region for heavy metal contamination due to intensive industrial activity, particularly leather processing. This research paper examines the ecotoxicological pathways and trophic transfer of heavy metals specifically Chromium (Cr), Cadmium (Cd), Lead (Pb), Copper (Cu), Zinc (Zn), Manganese (Mn), and Iron (Fe) from polluted irrigation water sources through agricultural soils and ultimately to food crops and animal feed consumed by local populations. The paper synthesizes findings from multiple studies conducted between 2010 and 2026, including soil, groundwater, and crop analyses across eight sites in Unnao. Results indicate that heavy metal concentrations in agricultural soils significantly exceed baseline values, with Pb ranging from 382.70 to 500.40 mg/kg and Cd from 79.60 to 293.80 mg/kg. This study provides a novel synthesis of zoological and ecotoxicological principles—specifically bioaccumulation, biomagnification, and species-specific susceptibility—to explain contamination patterns. Groundwater analysis reveals elevated Cr levels reaching 7.08 ± 1.42 mg/L in industrial-adjacent areas. Multivariate statistical analysis identifies cadmium, copper, lead, nickel, and zinc as anthropogenically sourced, primarily from tannery effluents and industrial discharge. The paper details the severe health consequences of this exposure, linking specific metals to cancers, cardiovascular disease, renal failure, and neurodevelopmental disorders in humans, as well as reproductive and physiological damage in livestock and wildlife. It concludes with policy recommendations for remediation, monitoring, and public health intervention based on a One Health approach.

DOI: https://doi.org/10.5281/zenodo.20393111

High Speed Floating Point Multiplication On FPGA Using Carry-look Ahead Adders

Authors: Dr.D.Satyanarayana, Dr.A.Ranganayakulu, pagadala Himanjali, Thota Lavanya, Sadu Snehalatha, Kappeta Madhavi

Abstract: Multipliers are crucial components for implementing algorithms for processing digital signals in hardware. When planning the architecture of a system as a whole, multiplier design is crucial. The floating-point format is preferred over the fixed-point representation for algorithms that need data with a dynamic range. However, designers have challenges with floating-point multipliers because to their enormous latency and space requirements. In this paper, we propose a spatially and temporally improved approach to floating point multiplication. The Mantissa multiplication, which is carried out by rapid tree multipliers, is the design bottleneck of the floating-point multiplier. The suggested architecture improved partial product reduction in mantissa multiplication by using Carry-lookahead adders as compressors, as opposed to half-adders and full-adders used in traditional Wallace or Dadda tree multipliers. Verilog HDL description is used to develop the FPGA and verify the suggested design. There was a 9.9 percentage point reduction in latency compared to Dadda and an 8.5% improvement compared to Wallace.

DOI: https://doi.org/10.5281/zenodo.20394359

IOT Based Automatic Wildlife Detection And Alert Systems For Farms

Authors: Mrs.N.Swarupa Rani, Mr.N.B.Jilani, Dr.A.Ranganayakulu, Anna. Suresh, Eraganaboyana Chandramohan, Metla Bala Venkata Narendra

Abstract: Using real-time detection, identification, and deterrent, the automated agricultural land security system is a smart AI-supported solution that aims to halt animal encroachment and safeguard crops. Because of their great sensitivity, distance-based sensors can keep a constant eye on the crops and their agent as they approach. An alarm buzzer notifies the farmer the moment it detects the animal. The AI system will then activate the YOLO object detection algorithm to determine the animal's species and trigger the transmission of a deterrent frequency tailored to that species in order to ward off any unwanted visitors. In the event that the animal persists or if there is evidence of manipulation, the farmer is alerted using GSM communication signals. Along the perimeter, we will also install an electric shock pulse system to beef up security against unwanted access. In order to analyze judgments based on sensor data instantaneously and provide timeless feedback for alarms, the microcontroller incorporates sense modes, deterrents, and communication modules. Thus, the system offers efficient, automated, and least disruptive animal deterrent, which should lessen crop damage and eliminate the need for human monitoring during certain times of the day. It offers a more long-term solution to the problem of wildlife management than the traditional approaches by using AI and an intelligent deterrence strategy to increase agricultural security and production.

DOI: https://doi.org/10.5281/zenodo.20394481

Low power 16-bit RISC Processor using vedic mathematics

Authors: Dr.D.Satyanarayana, Dr.A.Ranganayakulu, Maddasani Venkata Bala Krishnudu, Surarapu.V.V.Chandramohan Reddy, Challa Bhanu Prakash, Kola Bharath

Abstract: A 16-bit RISC processor with increased instruction execution capability via the use of Verilog and a Vedic multiplier design is the main focus of this study. For the purpose of design simulation, Vivado 2018.We utilize the 3D design suite. Applying the multiplier unit under the Arithmetic and logic unit (MAC) using the methods outlined in the Vedic Sutras is the primary objective of this study. Streamlining conventional computations in order to drastically reduce computing complexity is fundamental to Vedic mathematics. The data path allows information to move between different parts of the RISC processor, such as the memory, program counter, and register bank, and the control unit (which oversees the computer's operations) is another block that is part of the designed processor. The proposed RISC processor is quite simple and has the capability to carry out a grand total of fourteen commands. Compared to conventional ALU and MAC designs, this study successfully reduces power consumption in the Multiply-Accumulate (MAC) and arithmetic logic units (ALU) correspondingly. Both the Arithmetic Logic Unit (ALU) and the Multiply-Accumulate (MAC) have reduced delays as compared to their conventional counterparts. A 16-bit Vedic processor was born because of the gradual merging of the Vedic MAC and ALU with additional processing blocks. Compared to a regular CPU, this results in less delay and less power consumption. It follows that the most important characteristics of a well-designed CPU are a reduction in power consumption, an increase in operating speed, and smaller footprint. Verilog HDL, Vedic Mathematics, Von-Neumann architecture, Reduced Instruction Set Computer, and Sutras.

DOI: https://doi.org/10.5281/zenodo.20394625

Post Quantum Cryptography (PQC): Securing Data In The Age Of Quantum Computers

Authors: P. Ganga Bhavani, P. Nandini Lakshmi, V. Jahnavi, T. Harika, T. Sowmya

Abstract: The emergence of quantum computing is a serious threat to the principles of modern crypto-logical mathematics. The classical public-key cryptography, including Rivest-Shamir-Adleman (RSA), Elliptic Curve Cryptography (ECC), and Diffie-Hellman are very safe against classical adversaries yet become susceptible to quantum algorithms, including the Shor algorithm, which can easily solve the underlying mathematical problems. This weakness is jeopardizing confidentiality and integrity of the world-wide systems of communication, money transactions, and essential structures, and requires the immediate creation of safe alternatives. The current literature on post-quantum cryptography (PQC) has covered various classes of algorithm, such as lattice-based, code-based, multivariate schemes, and isogeny-based schemes. Although lattice-based approaches like Nth Degree Truncated Polynomial Ring Unit (NTRU) and FrodoKEM can be highly theoretically resistant, they can be computationally and memory-intensive. Cryptosystems based on code such as Classic McEliece are secure but not practicable because they have huge public key sizes. Multivariate schemes like Rainbow, and isogeny-based schemes like Super singular Isogeny Key Encapsulation (SIKE) have been attacked recently by cryptanalytic attacks, and broken completely. Such restrictions open the necessity to find more efficient and secure quantum-resistant solutions.This research study postulates a hybrid post-quantum cryptographic (PQC) system that combines CRYSTALS-Kyber for encryption with CRYSTALS-Dilithium for digital signatures both of which are approved through the National Institute of Standards and Technology (NIST) PQC standardization program. The model will be such that it ensures quantum resistance but at the same time it will be scalable, efficient and compatible with the existing infrastructures. The test outcomes indicate that the hybrid PQC model is 35-45 percent faster to encrypt and decrypt, signature performance is enhanced by up to 40 percent, and the throughput is 20 times higher (98.7 Mbps) than the same base PQC scheme including Nth Degree Truncated Polynomial Ring Unit (NTRU), McEliece and Rainbow. These enhancements affirm that the suggested model provides a viable trade-off between the security, speed, and efficiency in terms of key size, which makes it a viable and a sound base to ensure secure communication in the quantum age.

DOI: http://doi.org/10.5281/zenodo.20395331

Voice-Driven Detection Of Parkinson’s Disease Using Ensemble Machine Learning: A Comparative Study Of Acoustic Biomarkers”

Authors: Ishak Gauri, Himanshu Shrivastava, Irah Khan, Hritik Raj, Sohan Lal

Abstract: Parkinson's disease (PD) can be described as a debilitating disorder in which there is disruption to the dopaminergic pathway and associated with various motor dysfunctions. More importantly, one of the key but poorly exploited aspects of PD diagnosis is that vocal dysfunction occurs years before any motor symptoms. In this paper, a novel computational system for the diagnosis of PD through voice analysis is proposed. The proposed approach consists of feature extraction and the application of classification methods such as SVM, RF, KNN, and XG-Boost. Acoustic features including sustained phonation's jitter, shimmer, HNR, and Mel-Frequency Cepstral Coefficients (MFCC) are extracted and fed into machine learning algorithms. Linear kernel SVM provided the best result among all classifiers with an accuracy of 94.87% for training and 87.18% for testing with 195 data instances. Moreover, a web application for real-time PD diagnosis was developed with Flask backend and React frontend. It was shown that the biomarkers of voice signals are promising ways to non-invasively diagnose PD without much cost.

DOI: http://doi.org/10.5281/zenodo.20397202

CodeVenture – A Gamififed Learning Platform

Authors: Dr. Hemavathi C Purad, Tarun S, Mahammed Tousif, Mohammed Suhail, Santhosh Kumar B

Abstract: CodeVenture: A Gamified Learning Platform is designed to enhance the learning experience by integrating gamification techniques with interactive programming education. The platform aims to improve student engagement, motivation, and knowledge retention through game-based learning elements such as points, badges, leaderboards, levels, quests, and achievement tracking. Unlike traditional e-learning systems, CodeVenture provides an immersive and interactive environment where users can learn programming concepts through challenges, quizzes, coding missions, and real-time feedback. The platform incorporates AI-assisted learning support to guide students, personalize learning paths, and provide intelligent recommendations based on user performance. The system includes an intuitive and user-friendly dashboard that allows learners to track progress, monitor achievements, and visualize performance analytics effectively. The platform is developed using modern web technologies and follows an agile development methodology for iterative improvement and efficient deployment. The learning modules are designed to support multiple programming languages and provide hands-on coding practice through an integrated code editor and challenge-based exercises. Despite its advantages, the effectiveness of the platform depends on the quality of educational content, user participation, and adaptive learning mechanisms. Excessive gamification may distract learners from educational objectives, and different users may respond differently to reward-based learning strategies. Additionally, implementing AI-based personalization and real-time analytics may introduce computational and scalability challenges. The primary objective of this work is to develop an interactive gamified learning system that enhances programming education through engaging and adaptive learning techniques. The platform further aims to provide personalized learning experiences, performance visualization, and intelligent feedback mechanisms to improve student understanding and participation. CodeVenture combines gamification principles, AI-assisted learning support, and interactive web technologies to create a scalable, motivating, and effective educational environment. Compared to conventional learning systems, the proposed platform improves learner engagement, encourages continuous skill development, and creates an enjoyable and productive learning experience for students from both technical and non-technical backgrounds.

AI And IoT Based Pest Detector: An IoT-Enabled Multi-Sensor And AI-Integrated System For Crop Infestation Detection And Advisory Using CNN And Generative AI

Authors: Vedarsh Lokhande, Omkar Holkar, Om Gadilkar, Kshitij Nawale, Tirupati Hale, Dr. N.P Bhone

Abstract: Pests and diseases on crops are a crucial challenge impacting agriculture productivity, causing global losses up to 20%-40% annually [1]. In this paper, we propose AI and IoT Based Pest Detector (IPD), an IoT-AI integrated system capable of real-time crop infestation detection and intelligent advisement generation. It integrates a hardware node based on ESP32 (equipped with various environmental sensors: DHT22(temperature/humidity), capacitive soil moisture sensor, leaf wetness sensor, MQ-135 gas sensor, KY-038 sound sensor, and OV2640 camera module) with a pipeline that blends deep learning for image analysis and large language model (LLM). The training of a custom Convolutional Neural Network (CNN) to recognize 38 different classes of plant diseases based on the PlantVillage dataset [2] shows an accuracy of 97.07% on training data and 95.08% on validation data over 10 epochs. When a leaf image is uploaded, an inference pipeline first classifies if the image falls in the 38 recognized categories; if so, it applies the trained CNN for classification and a Gemini API [12] to generate an advisory (infestation severity, organic, chemical treatment, prevention strategies and helpful agricultural numbers); if not, Gemini would give out a direct AI analysis and suggestion. All the detected information are presented on a web dashboard made by Streamlit and simultaneously sent to a local OLED screen. Our proposed system is an all-in-one and cost-effective solution, expandable to both small farms and precision farming contexts.

Design And Implementation Of A 50 TPD Green Hydrogen Production Facility Using Alkaline Electrolysis Technology For LCOH < $2/kg

Authors: Pranav Gupta, Arjun Dhama, Rukminesh Tiwari, Manaswini Pandela

Abstract: This paper presents the design and implementation of a 50 tons per day (TPD) green hydrogen production facility utilising Alkaline Electrolysis Cell (AEC) technology, aimed at achieving a levelized cost of hydrogen (LCOH) below $2/kg. AEC technology was selected for its long lifespan, low capital and operational expenditures, and high technology readiness level. The facility is strategically located in Gujarat, India, leveraging the region's abundant renewable energy resources and supportive government policies. The plant operates with a feedstock of 450 kL/day of deionized water sourced from the Narmada River and employs a round-the-clock renewable energy model combining 300 MW of solar, 200 MW of wind, and 140 MW of hydro power, supplemented by a 10 MW Battery Energy Storage System (BESS). The produced green hydrogen is intended for use in the fertilizer and petrochemical industries, replacing grey hydrogen and natural gas, and as an energy carrier. The project is projected to offset 274,000 tonnes of CO₂ annually, highlighting its significant environmental benefits. This paper discusses the technical design, economic feasibility, and potential impact of the facility on the green hydrogen economy.,

DOI: http://doi.org/10.5281/zenodo.20400032

Resource Allocation In Cloud Computing: A Critical Analysis Of Profit-Driven Decision-Making Models

Authors: Dr.D.Balasubramanian, Dr. Megala.R

Abstract: Cloud computing has become a dominant paradigm for delivering scalable and on-demand computing services. Efficient resource allocation is essential for maintaining service quality, maximizing infrastructure utilization, and improving provider profitability. Most contemporary cloud resource management systems employ profit-driven decision-making models that prioritize economic gains and operational efficiency. However, these approaches often neglect fairness, energy efficiency, sustainability, and Quality of Service (QoS) guarantees. This paper critically analyzes the limitations of profit-oriented resource allocation strategies in cloud environments and evaluates their impact on SLA compliance, energy consumption, and user satisfaction. A sustainable multi-objective resource allocation framework is proposed to integrate profit optimization with QoS assurance and energy-aware scheduling. Experimental analysis using CloudSim demonstrates that hybrid multi-objective models outperform purely profit-driven approaches in terms of SLA reduction, energy efficiency, and workload fairness. The proposed framework provides a balanced and sustainable solution for future cloud infrastructures.,

DOI: http://doi.org/10.5281/zenodo.20404369

Phytochemical, Pharmacological, And Ethnobotanical Evaluation Of Aloe Vera (L.) Burm. F.: A Comprehensive Study On Its Medicinal Applications And Therapeutic Potential

Authors: Archana Kumari, Munchun Kumari, Md. Faizan Alam, Divakar, Amrita Kumari, Dr. Balwant Singh

Abstract: Aloe vera (L.) Burm. f., commonly known as Ghritkumari, is an important medicinal plant extensively utilized in traditional and modern healthcare systems because of its remarkable therapeutic and pharmacological properties. The present study was undertaken to evaluate the botanical characteristics, phytochemical composition, pharmacological significance, and ethnobotanical applications of Aloe vera through integrated laboratory analysis, microscopic examination, ethnobotanical survey, and literature evaluation. Authenticated plant materials collected from different medicinal plant centers were subjected to sequential solvent extraction and phytochemical screening using standard analytical methods. The results revealed the presence of diverse bioactive compounds including anthraquinones, flavonoids, tannins, saponins, sterols, phenolic compounds, polysaccharides, glycosides, proteins, and mucilage. Methanolic and aqueous extracts exhibited the highest phytochemical richness and biological potential. Microscopic studies confirmed characteristic anatomical features such as epidermal tissue, chlorenchyma, latex-containing pericyclic cells, and mucilage-rich parenchymatous gel tissue. Ethnobotanical investigations demonstrated extensive traditional use of the plant in treating burns, wounds, skin diseases, gastric disorders, constipation, diabetes, inflammation, and hair problems among local healers and herbal practitioners. Pharmacological evaluation highlighted significant wound-healing, anti-inflammatory, antioxidant, antimicrobial, antidiabetic, immunomodulatory, and anticancer activities primarily associated with compounds such as acemannan, aloin, aloe-emodin, vitamins, enzymes, and phenolic antioxidants. The study validates the scientific basis of traditional medicinal uses of Aloe vera and emphasizes its considerable pharmaceutical, nutraceutical, cosmetic, and commercial importance. Furthermore, the findings highlight the need for standardized extraction procedures, dosage optimization, and long-term clinical studies to ensure safe and evidence-based therapeutic utilization.

DOI: http://doi.org/10.5281/zenodo.20404560

Intelligent Load-Aware Seat Allocation for Railway Coaches: A Greedy Heuristic Approach with Simulation-Based Validation

Authors: Shweta Arun Singh, Tejal Vinod Apraj, Dr. Jasbir Kaur, Suraj Kanal, Sandhya Thakkar

Abstract: Efficient seat allocation in large-scale railway reserva- tion systems remains challenging due to uneven passenger weight distribution across coaches. Existing platforms such as IRCTC employ deterministic sequential allocation strategies that prioritise berth preferences but do not explicitly optimise for real-time load balancing. This paper presents the Intelligent Load-Aware Seat Allocation (ILASA) framework, a greedy minimum-load heuristic that dynamically assigns seats by incorporating pas- senger weight, age, journey segment, and berth preference into a single multi-constraint decision process. The framework is evaluated through discrete-event simulation using synthetic passenger data calibrated against published Indian Railways demographic statistics. Across 100 independent simulation runs per experimental condition, ILASA achieves a 55.1% reduction in inter-coach load imbalance (standard deviation of coach loads) and a 22.0% improvement in average seat utilisation under peak occupancy, compared against sequential and random allocation baselines. Mean allocation latency of 12.4 ms satisfies real-time booking requirements. We explicitly acknowledge the limitations of this study: reliance on synthetic rather than operational railway data, a simplified passenger weight model that does not yet account for luggage, group travel constraints, or dynamic passenger movement, and the heuristic nature of the algorithm which has not been benchmarked against metaheuristic or mathematical programming alternatives. A web-based prototype demonstrates the feasibility of real-time visualisation. This work provides a foundation for future research incorporating real-world operational data, more sophisticated optimisation techniques, and comprehensive passenger acceptance studies.

DOI: https://doi.org/10.5281/zenodo.20406082

Academic Result Management System Using Django

Authors: Mudupu sravya, Mujja vamshi, Putti srinath

Abstract: — The Academic Result Management System is a web-based application designed to efficiently manage and track student academic performance, including courses, grades, and related data. It simplifies administrative tasks by providing a centralized platform for teachers and administrators to enter, organize, and retrieve information, while offering modules like student, course, result, and report management. The system also enables students to securely access their results through a user-friendly interface. By reducing manual errors and improving efficiency, it supports better decision – making and enhances overall academic management in educational institutions.

Thinkbotting: Are Students Outsourcing Their Ability To Think? Cognitive Dependency, AI-Assisted Learning, and the Future of Student Thinking

Authors: Deepika Sikha

Abstract: The rapid integration of generative artificial intelligence into education is transforming how students learn, complete assignments, and engage with knowledge. While AI-assisted tools offer accessibility, efficiency, and personalized support, they have also introduced a growing behavioural phenomenon termed thinkbotting—the tendency of students to outsource cognitive effort, reasoning, reflection, and problem-solving to AI systems rather than engaging in independent thinking. This paper explores the educational implications of excessive AI dependency among school and university learners. Drawing from cognitive psychology, constructivist learning theory, educational sociology, and emerging classroom observations, the article examines how overreliance on AI-generated responses may weaken critical thinking, creativity, academic resilience, and intellectual autonomy. The paper argues that education systems risk producing technologically efficient but cognitively passive learners if reflective and inquiry-based learning practices are not preserved. The article further discusses the changing role of teachers, assessment systems, and ethical learning in AI-mediated classrooms. It concludes by advocating for balanced AI integration models that position artificial intelligence as a cognitive support tool rather than a substitute for human thought, curiosity, and intellectual struggle.

AI Virtual Keyboard with Gesture Recognition Using Python: A Real-Time Hand Tracking Approach with MediaPipe and OpenCV

Authors: Ms. Sharvari Masurkar, Ms. Samruddhi Kondekar, Dr. Jasbir Kaur, Assistant Professor Ms. Sandhya Thakkar

Abstract: The rapid advancement of artificial intelligence and computer vision has significantly transformed the field of Human- Computer Interaction (HCI). Physical input devices such as keyboards and mice, while efficient, require direct tactile contact, which may be undesirable in contexts demanding heightened hygiene, accessibility, or convenience. This paper presents the design, implementation, and evaluation of an AI Virtual Keyboard system that enables touchless typing through real-time hand gesture recognition. The proposed system leverages OpenCV for video capture and image processing, and MediaPipe for detecting 21 hand landmarks that are subsequently translated into keystrokes using Euclidean distance-based gesture interpretation. A comprehensive user study involving 30 participants evaluated the system on accuracy, typing speed, and usability. Experimental results demonstrate a gesture detection accuracy of 92–95%, precision of 93%, and an average typing speed of 12–15 words per minute (WPM) for novice users and up to 18–20 WPM for experienced users. The system was tested under varying lighting conditions (150–600 lux) and achieved consistent performance with an average response time of 1.0–1.5 seconds. User feedback indicates 80% of participants found the system easy to learn within minutes. While slower than traditional physical typing (40– 60 WPM), the proposed system offers a viable touchless alternative for public kiosks, healthcare environments, and accessibility applications.

Bridging Authentic Leadership And Resilience: The Mediating Role Of Indian Knowledge System Based Coping Skills

Authors: Mrs. Pujalin Mishra, Dr. Manoj Kumar Sethi

Abstract: In the context of increasing workplace complexity, stress, and uncertainty, understanding how employees develop resilience has become a significant concern in organizational research. This qualitative study explores the role of coping skills, grounded in both the Indian Knowledge System (IKS) and modern psychological perspectives, as a mediating mechanism in the relationship between authentic leadership and employee resilience. Authentic leadership, characterized by self-awareness, ethical conduct, relational transparency, and balanced decision-making, is considered a critical influence on employees’ psychological and Behavioural outcomes. However, the processes through which it contributes to resilience require deeper, context-sensitive exploration. Adopting an interpretive qualitative approach, this study draws on in-depth interviews and thematic analysis to capture employees lived experiences within organizational settings. It investigates how individuals perceive authentic leadership behaviours and how such leadership fosters the development of coping strategies. Particular attention is given to coping practices derived from the Indian Knowledge System, including mindfulness (Dhyāna), equanimity (Samatva), self-awareness (Ātma-bodha), and selfless action (Karma Yoga), alongside modern coping strategies such as emotional regulation, cognitive reframing, and problem-solving. The findings are expected to reveal that authentic leadership creates a supportive and trust-based environment that encourages self-reflection, meaning-making, and emotional balance. These conditions facilitate the adoption of both traditional and modern coping mechanisms, which in turn strengthen employees’ ability to adapt, recover, and grow in the face of adversity. The study contributes to the literature by providing a nuanced understanding of the interplay between leadership, culturally embedded coping processes, and resilience. By integrating indigenous knowledge with contemporary organizational theory, this research offers a holistic framework for enhancing resilience and highlights the importance of culturally relevant coping strategies in leadership development and organizational practice.

DOI: http://doi.org/10.5281/zenodo.20407933

The JJM Utilization Performance Index (JUPI): A Multi-Dimensional Framework For Assessing Rural Water Governance Under India’s Jal Jeevan Mission

Authors: Mustaq Shaikh, Farjana Birajdar

Abstract: India's Jal Jeevan Mission (JJM), launched in August 2019 with the goal of providing Functional Household Tap Connections (FHTC) to every rural household by 2024, represents one of the largest rural infrastructure programmes in global history. This paper presents a comprehensive governance case study of JJM implementation in Solapur District, Maharashtra—a water-scarce, semi-arid region characterised by basaltic terrain and recurrent drought conditions—analysing progress across six performance dimensions: FHTC household coverage, Har Ghar Jal (HGJ) village certification, Jal Seva Aankalan (JSA) water quality assessments, institutional coverage of schools and Anganwadi Centres (AWCs), eGramSwaraj digital platform onboarding, and scheme financial completion. Using official data sourced from the District Water and Sanitation Mission (DWSM) as of May 2026, the study reveals that Solapur has achieved near-universal FHTC coverage of 99.90% (5,76,668 of 5,77,245 rural households), with 8 of 11 administrative blocks attaining 100% connection rates. Despite this physical infrastructure success, critical implementation gaps persist: HGJ village certification stands at only 59.3% (662 of 1,116 villages), scheme financial completion is critically low at 11.7% (178 of 1,525 schemes), AWC tap water coverage averages 59.4% with extreme block-level variation (Madha: 7.4%; Karmala: 100%), and JSA assessments remain incomplete in 27 villages across 9 blocks. The study introduces the JJM Utilization Performance Index (JUPI), a composite governance metric integrating all six dimensions, revealing a district average of 84.0 (range: 75.9–93.0), with Akkalkot (93.0), Sol. North (91.7), and Sol. South (90.8) as high performers and Sangola (76.6) and Mangalvedhe (75.9) requiring targeted interventions. The findings demonstrate that infrastructure delivery, while necessary, is insufficient for sustainable water security; effective governance, community ownership, financial accountability, and institutional equity are equally essential. Policy implications for the broader national JJM programme are discussed.

DOI: http://doi.org/10.5281/zenodo.20408692

Ai-Based Appointment Scheduling Assistant

Authors: Ram Bhatt, Dr. Raj Kumar, Kunal Kumawat, Karan Kumar, Gaurav Prakash

Abstract: Moreover, with the rapid growth of digital technologies, the need for smart, efficient, and scalable appointment scheduling solutions is increasing across service-based industries such as healthcare, education, and professional services. Traditionally, appointment scheduling systems were driven only by pure manual coordination or fixed forms-based scheduling solutions. Some of the challenges connected with traditional appointment scheduling solutions include conflict scheduling, late confirmations, poor resource utilization, a heavy administrative burden, and issues with user experience, especially within industries where growing demands for services have presented a challenge to efficient operations and service delivery. Recent developments in AI and web technologies offer new scopes for improvement in appointment management systems through automated decision-making by AI-based intelligent systems. In this context, this research work proposes an AI-Based Appointment Scheduling Assistant built on the MERN technology stack-MongoDB, Express.js, React.js, and Node.js-associated with Natural Language Processing and rule-based AI technology. The proposed work enables appointment scheduling through natural language statements, hence offering a more user-friendly and interactive experience to clients. The project makes use of NLP techniques for extracting critical appointment details such as date, time, and purpose, while a rule-based AI engine checks appointment constraints and rules for compliance with pre-defined business rules. Implementation includes the authentication and authorization process through JSON Web Tokens for ensuring security aspects related to accessing the system and ensuring the integrity of data. The system uses MongoDB, a flexible and scalable storage of all data related to appointments that will be done on the system. Experimental verification on the proposed system proves the improvements in scheduling accuracy, conflict resolution, human intervention, and system response time over the conventional scheduling method. Experimental results suggest that the use of NLP intelligence and expertise systems on a contemporary full-stack solution can greatly optimize the efficiency of a system. The proposed work also facilitates a cost-effective, intelligent, and efficient way for appointment scheduling and lays the background for performing improvisations in the field of prediction, machine learning, and multiple language support.

DOI: https://doi.org/10.5281/zenodo.20409577

Demand Forecasting Using Deep Learning For Resilient And Agile Supply Chain Networks

Authors: Dhanusha Mol K P, Dr. S. Ilankumaran

Abstract: Supply chains around the world have been becoming more prone to various forms of disruptions from pandemics to geopolitical tensions, and it has revealed the inadequacy of existing forecasting methods. Successful demand forecasting is the foundation of a resilient supply chain, allowing for strategic inventory management and capacity planning. This study develops an advanced demand forecasting system based on deep learning that leverages a Temporal Fusion Transformer (TFT) and multiple sources of external data such as weather information, economic statistics, social media trends, and supply chain disruptions. Using the data collected for five years (2019-2025) from the multinational retail supply chain that amounts to 50 million SKU-location-weeks, the TFT model demonstrates WAPE = 12.4% when making forecasts for four weeks ahead, surpassing other forecasting models (ARIMA – 24.8%, XGBoost – 18.2%, and LSTM – 15.6%). Moreover, the developed system features a unique disruption-aware training process that increases forecast precision during disruptions by 28%. When tested in conjunction with a multi-echelon inventory management system, the forecasting system was able to cut the amount of safety stocks by 19% and improve on-time delivery performance by 31%.

AI Vision Framework For Wildlife Injury Detection And Rescue Alert In Dense Forest Environments

Authors: S. Nithyananth, SatheeshKumar S, Sathya Prakash. S, Sasivanan S, Sivaram S, Somiya M

Abstract: Illegal poaching, road accidents, and natural calamities in dense forests have led to the infliction of injuries on wildlife, which has resulted in extended periods of suffering and even death prior to any human intervention. Manual patrols are difficult, ineffective, and unsafe. In this paper, we propose an innovative AI Vision Framework for Wildlife Injury Detection and Rescue Alert (WIDRA) that leverages edge-based camera traps, unmanned aerial vehicles (UAVs), and deep learning algorithms to detect and classify injured animals in dense forests. The framework consists of three key components: (1) YOLOv8 for detection and classification of animals, (2) Temporal Convolutional Network (TCN) with attention mechanism to assess the level of injury through movement and postural analysis, and (3) LoRaWAN-based system for geotagged rescue notifications. Evaluated in simulated dense forest environments across two wildlife sanctuaries, the proposed system has attained 94.2% accuracy in animal detection, 89.6% sensitivity for injury classification, and shortened the rescue response time from 18 hours (with manual patrols) to 45 minutes.

DOI: https://doi.org/10.5281/zenodo.20415465

Permission-Based Android Malware Classification Using Genetic Algorithm

Authors: Anukeerthana D, Sathiya Shree S, Dr. P. Jeyanthi

Abstract: The exponential growth of Android applications has made the platform highly attractive to cybercriminals, who exploit its open-source nature and permission model to distribute malicious software. Traditional signature-based detection methods are ineffective against zero-day and obfuscated malware, while dynamic analysis approaches are computationally expensive and carry execution risks. This paper proposes the Intelligent Android Malware Detector, a web-based system that performs safe static analysis on Android Package Kit (APK) files. Key features, primarily requested permissions and application metadata, are extracted without executing the application. A Genetic Algorithm (GA) optimizes the feature set by selecting the most discriminative permissions and eliminating redundant ones, reducing dimensionality and improving model efficiency. The optimized features are then fed into an Artificial Neural Network (ANN) that learns complex patterns and outputs a malware probability score for nuanced risk assessment, achieving a classification accuracy of 92.26%. To enhance usability and awareness, the system incorporates a hybrid AI chatbot for explanatory support and a real-time threat intelligence module that aggregates cybersecurity news. Implemented using the Flask framework, the proposed solution offers a proactive, user-friendly, and scalable approach to Android malware detection, addressing key limitations of existing systems while promoting cybersecurity education.

DOI: https://doi.org/10.5281/zenodo.20415604

Data-Driven Pricing Strategies In Online Retail Platforms For Revenue Maximization

Authors: Dr.S.Subalakshmi, Dr. S. Udhaya

Abstract: The rise of the internet has made pricing a much more dynamic process than ever before. In this paper, we propose an end-to-end solution to implement data-driven pricing optimizations through an integration of demand forecasting, price elasticity, and pricing optimization components. We build a multi-horizon forecasting model leveraging TFTs for accurate future demands forecasts, a product-level price elasticity model based on Bayesian structural time-series models, and a revenue maximization optimization engine to find optimal prices. Our methodology, trained on historical transaction data consisting of 5 million sales transactions for 10,000 SKUs over a 3-year period (2023-2025) achieved a revenue lift of 12.4% (p<0.01) in an A/B test versus two benchmark pricing methods such as cost-plus pricing (4.2% lift) and competitor-based pricing (6.1% lift). The paper concludes with a discussion of implementation challenges and practical guidelines for deploying algorithmic pricing in competitive retail environments.

DOI: https://doi.org/10.5281/zenodo.20415916

Fraud Detection In Signature Verification Using Advanced Image Processing Techniques For Real-Time Authentication

Authors: P. Anusha, B. Niharika, A. Amitha, G. Anshuman

Abstract: This research focuses on developing a feasible and efficient solution for verifying handwritten signatures using advanced image processing techniques. The study is specifically limited to signatures that involve static inputs and outputs, meaning that it does not take into account dynamic elements such as writing speed or pressure. To identify the most suitable classifier for accurate signature verification, multiple machine learning models were explored, including the Multinomial Naïve Bayes Classifier (MNBC), Bernoulli Naïve Bayes Classifier (BNBC), Logistic Regression Classifier (LRC), Stochastic Gradient Descent Classifier (SGDC), and Random Forest Classifier (RFC). Each classifier was trained and evaluated using a publicly available signature dataset to ensure consistency and reliability in performance measurement. After rigorous testing, the Random Forest Classifier (RFC) demonstrated the highest accuracy, achieving a score of approximately 0.99. This suggests that RFC is the most effective model among those tested for distinguishing between genuine and forged signatures. On average, the system has proven to be highly successful in verifying signature images with a significant level of accuracy. The results of this study indicate that machine learning-based approaches, particularly RFC, can provide a reliable method for signature authentication, which could be beneficial in various real-time applications such as banking, legal document verification, and identity authentication.

DOI: https://doi.org/10.5281/zenodo.20423387

Machine Learningapproaches for Estimating Drinking Water Safety: Assessing Human Consumption Suitability

Authors: K. Vigneshwar, A. Vedhika, B. Sai Teja, B. Josmitha

Abstract: Drinking Water Supply (DWS) systems are among the most essential and sensitive infrastructures required for maintaining urban life and public health across the world. In Europe, rapid population growth combined with aging and obsolete water supply infrastructure has created significant challenges in ensuring safe and continuous water distribution. Maintaining high water quality standards is critical not only for providing clean water for daily consumption but also for preventing health hazards caused by contamination. Traditional water quality monitoring methods mainly rely on periodic laboratory testing of parameters such as pH, turbidity, dissolved oxygen, and bacterial content. However, these testing procedures generally require 24–48 hours to produce results, creating a delay in identifying contamination and increasing the risk of bacterial spread within the water distribution network. To address these issues, this study proposes an Exploratory Data Analysis (EDA) based model for water quality assessment and prediction. The proposed model considers two major dimensions: water quality parameters and water quality score. Furthermore, machine learning techniques are applied to predict water quality changes within the DWS system. In this research, the Random Forest algorithm is implemented using PyCaret for efficient model development and analysis. A case study was conducted on an industrial water supply system to evaluate the model’s effectiveness. The preliminary results demonstrate that the proposed approach can successfully analyze and predict water quality conditions, helping authorities improve monitoring efficiency and reduce response time to contamination risks.

DOI: http://doi.org/10.5281/zenodo.20425611

Harnessing Ai to Gauge Citizen Emotions in Smart Cities

Authors: K. Vigneshwar, T. Praveen Reddy, T. Mahesh, S. Pranay

Abstract: Over the past decade, smart city applications have gained significant attention in industrial informatics. However, little attention has been given to perceiving the emotions and perceptions of citizens who have a direct impact on smart city initiatives. We propose the use of publicly available, abundant social media conversations that contain contextual information encompassing citizens' emotions and perceptions which could be considered to provide the means to feel the ‘emotional pulse’ of a city. We propose an automated AI-based observation framework to detect the emergence of public emotions and negativity in conversations. We evaluated the applicability of the framework using 29,928 social media conversations towards the much debated topic of self-driving vehicles which will become increasingly relevant to smart cities. The patterns and transitions of citizens’ collective emotions were modeled using the NLP and Markov models while the negativity (toxicity) in conversations was evaluated using a deep learning based classifier. The framework could be adopted by industry leaders and government officials for smart observation of citizen opinions to improve security, communication, and policymaking.

DOI: http://doi.org/10.5281/zenodo.20426043

AI Based Essay Analysis System

Authors: Manish Raj Pandey, Ranjeet Kumar, Shashank Singh, Abhishek Chauhan, Dr. Yashveer Singh

Abstract: — In the digital age, students and educators face increasing difficulty in evaluating written content efficiently, as a massive amount of essays, assignments, and reports are produced every day. It is becoming more challenging to review each document manually and extract meaningful insights due to the growing volume of academic writing. As a result, there is a rising demand for automated systems that can analyze essays quickly and accurately, helping users understand key points without spending excessive time and effort. The goal of this project is to present a comprehensive study of modern techniques and tools used for AI-based essay analysis. We aim to explore different approaches, ranging from basic rule-based methods to advanced machine learning and deep learning models, and discuss their strengths, limitations, and possible applications. By evaluating and comparing stateofthe-art techniques, we aim to highlight current trends and challenges in this domain, offering valuable insights for future development. In summary, this project provides an in-depth review of advanced methods for automated essay evaluation, focusing on grammar checking, coherence detection, sentiment understanding, and scoring. Our objective is to enhance awareness of the opportunities and challenges in automated essay analysis and inspire further research and innovation, enabling students, educators, and institutions to evaluate written content more effectively and efficiently.

DOI: https://doi.org/10.5281/zenodo.20426074

Industrial Applications Of Microbial Enzymes: A Comprehensive Review

Authors: Himadri Sharma

Abstract: Microbial enzymes have become essential biocatalysts in modern biotechnology because of their high specificity, catalytic efficiency, and eco-friendly nature. Enzymes derived from bacteria, fungi, yeast, and actinomycetes are extensively utilized in food processing, pharmaceuticals, agriculture, textiles, detergents, leather, paper, and biofuel industries. Compared to conventional chemical catalysts, microbial enzymes function under mild environmental conditions and reduce industrial pollution. Recent developments in molecular biology, protein engineering, recombinant DNA technology, and fermentation processes have significantly enhanced enzyme production and stability. This review discusses the sources, production methods, classification, industrial applications, advantages, limitations, and future prospects of microbial enzymes in industrial biotechnology.

DOI: https://doi.org/10.5281/zenodo.20426583

End-to-End CNN-Based System for Human Detection in Fire Scenes Deployed via a Flask Web Application

Authors: Dr. Rajkumar, Md Nasir Hussain, Harsh Kumar, Shashikesh Kumar, Aman Mehra, Shubham Mahto, Irah Khan

Abstract: Fire emergencies continue to claim thousands of lives and cause enormous economic damage every year across residential, commercial, and industrial settings. While conventional fire alarm systems based on smoke or heat sensors remain the dominant approach in modern buildings, they are inherently reactive, slow to respond in large spaces, and entirely blind to the presence and location of human occupants inside hazardous areas. This paper presents a complete, deployment-ready fire and human detection system built around two complementary deep learning models: a custom binary Convolutional Neural Network (CNN) trained to classify fire and non-fire scenes, and the YOLOv8 Nano one-stage object detector configured to localise human occupants in real time. The outputs of both models are fused inside a Decision Engine that generates one of four contextual threat levels — SAFE, PERSON DETECTED, FIRE ONLY, or HIGH RISK — along with rich visual annotations including bounding boxes, hazard overlays, and severity banners. The system is delivered through a Flask REST backend hosted on Hugging Face Spaces and a React-Vite single-page application that streams live webcam frames entirely through the browser using the HTML5 getUserMedia and Canvas APIs, thereby removing the need for server-side camera access. On the evaluation set, the CNN achieves 96.38 % accuracy, 97.20 % precision, 95.50 % recall, and an F1-score of 96.34 %, while the YOLOv8 Nano detector attains a mean Average Precision at IoU threshold 0.50 (mAP50) of 89.2 %. Frame-skipping and lazy model initialisation raise real-time throughput from 5 FPS to 25 FPS and cut CPU utilisation from 98 % to 35 % on a standard cloud CPU. The results demonstrate that a carefully engineered lightweight architecture, even without GPU acceleration, can serve as a practical, scalable, and cost-effective intelligent safety monitoring solution.

DOI: https://doi.org/10.5281/zenodo.20426948

</p

Semantic Video Discovery Using Deep Feature Fusion And Automated Metadata Generation

Authors: Usha Dhankar, Komal Khatak, Dr. Sweety

Abstract: The exponential growth of video data across domains such as surveillance, aerospace, and digital media has created a significant challenge in efficient content retrieval. Traditional approaches based on manual tagging and low-level visual features fail to capture the contextual semantics of video content. This paper proposes a semantic video discovery framework that integrates deep feature fusion with automated metadata generation. Visual features are extracted using deep learning models such as YOLO and Segment Anything Model (SAM), while textual features are derived using Natural Language Processing (NLP) techniques including FastText and Named Entity Recognition (NER). The fusion of visual and textual embeddings enables context-aware retrieval and improves semantic understanding of video content. Experimental results demonstrate enhanced accuracy, precision, and retrieval efficiency compared to traditional methods.

DOI: https://doi.org/10.5281/zenodo.20438784

Atezhare: A Session-Based Cloud Integrated File Sharing Mobile Application

Authors: Basil V Mathew, Abineeth Tm, Inbarasu M, Koushick Ganapathi Venkat M, Dr. Rajesh Babu

Abstract: The rapid growth of mobile technology has created significant demand for seamless, secure, and efficient file sharing solutions. Traditional file sharing methods often rely on internet connectivity, cloud servers, or third-party intermediaries, raising concerns about privacy, speed, and data security. This paper presents Atezhare, a session-based Hy (H-(H-P2P)) file sharing mobile application that enables direct device-to-device file transfer without the need for internet access or external servers. The application is built natively for Android and utilizes a dual-mode pairing mechanism comprising QR code scanning and a 6-digit alphanumeric session code to establish cryptographically bounded, time-limited connections between devices. Atezhare incorporates a cloud-integrated session management backend powered by Spring Boot and hosted on Render, with Supabase providing persistent session metadata storage. Experimental evaluations demonstrate that the system achieves reliable Hybrid-peer-to-peer connectivity with connection establishment times averaging 3–5 seconds, transfer speeds comparable to cloud-based alternatives under local wireless conditions, and enhanced data privacy due to the elimination of third-party server routing. The paper presents the system's architecture, module description, security mechanisms, implementation details, and comparative evaluation against existing file sharing solutions.

DOI: http://doi.org/10.5281/zenodo.20438779

Distributed Messaging System { Kafka }

Authors: Ricky Das, Vipin Kumar Dhiman

Abstract: Nowadays, log processing is very important for internet-based companies. In this paper, we talk about Kafka — a distributed messaging system we created to collect and deliver a large amount of log data with very low delay. Kafka uses some ideas from old log systems and messaging tools, and it works well for both offline and online data reading. While making Kafka, we took some different but useful steps to make it fast and scalable. Our tests showed that Kafka works better than two other famous messaging systems. We are already using Kafka in real-life systems, and it handles hundreds of gigabytes of new data every day.

Gesture Controlled Virtual Mouse Using Computer Vision

Authors: Mr.P. Loganathan, Beula C, Mathumitha M, Kavipriya K, Vishnupriya K

Abstract: The rapid advancement of computer vision and machine learning has opened transformative avenues for Human-Computer Interaction (HCI). Traditional input devices such as the mouse and keyboard impose physical constraints that make them inaccessible to users with motor impairments and impractical in sterile, hazardous, or presentation contexts. This paper presents the design and implementation of a Gesture Controlled Virtual Mouse system that enables touchless, real-time cursor control through hand gestures captured by a standard RGB webcam. The proposed system leverages Google's MediaPipe Hands framework for 21-point three-dimensional hand landmark detection, OpenCV for video acquisition and visual feedback rendering, and PyAutoGUI for system-level mouse event execution. A lightweight, rule-based gesture classifier interprets finger configuration vectors to map specific hand postures to mouse actions including cursor movement, left- click, right-click, double-click, scroll, and drag operations. Coordinate smoothing via an exponential moving average filter mitigates cursor jitter, and a time-based cooldown mechanism prevents unintended repeated click events. Experimental evaluation conducted across three lighting conditions on 100 trials per gesture demonstrates an overall gesture recognition accuracy of 92.8%, with cursor movement achieving 96% accuracy. The system operates at 26 frames per second on commodity hardware, yielding an end-to-end processing latency of approximately 38 milliseconds. A System Usability Scale evaluation with 15 participants yielded a score of 76.3 out of 100, indicating good usability. The proposed system presents a cost- effective, hardware-independent alternative to conventional mouse input and holds significant potential for accessibility, healthcare, education, and industrial applications.

DOI: http://doi.org/10.5281/zenodo.20438891

Career Intelligence and Roadmap Recommendation System

Authors: Poornima K, Praveen A, Abinaya M P, Phavinpriya M, Dr. M. Rajesh Babu

Abstract: Fresh graduates often face significant challenges in identifying the right career path due to unclear role definitions, rapidly evolving job market demands, and limited access to personalized guidance. This paper proposes a Career Intelligence and Roadmap Recommendation System that leverages data analytics and Natural Language Processing (NLP) to provide intelligent, data-driven career guidance for students and fresh graduates. The system collects real-world job posting data from publicly available job portals to analyze current industry requirements, salary trends, and skill demands. It performs a comprehensive skill gap analysis by comparing user-provided skills with market-required competencies for selected job roles. Based on this analysis, the system generates a structured, step-by-step career roadmap that dynamically updates as users acquire new skills. The platform integrates visual analytics such as demand charts and salary trend graphs to help users understand industry expectations. The proposed system aims to reduce skill mismatches, improve job readiness, and enhance overall employability among students and fresh graduates.

DOI: http://doi.org/10.5281/zenodo.20439291

Mental Health Tracking System Using Java, React.js, and MySQL

Authors: Siddhi Umesh Kurhade, Vedant Santosh Kokane, Bhumika Sunil Temkar, Payal Ashok Ware, Prof. Khatal K. B.

Abstract: Mental health disorders are rapidly increasing among students and working professionals due to stress, anxiety, workload imbalance, and lifestyle challenges. Traditional mental health monitoring approaches mainly depend on counseling sessions and manual observation, which often lack continuous tracking and accessibility. This research presents the design, development, implementation, and result-based evaluation of a full-stack Mental Health Tracker Application developed using Java Spring Boot, React.js, and MySQL. The system enables users to record daily moods, maintain emotional journals, monitor behavioral patterns, and analyze emotional trends using dashboard analytics. Experimental anal-ysis was conducted using 50 participants over a testing duration of 14 days. The system demonstrated 92% user satisfaction, 95% reliability, 98% uptime, and improved user engagement in emo-tional self-monitoring activities. Performance testing showed low API response time and efficient database operations. The research findings indicate that full-stack web technologies can effectively support digital mental health monitoring and emotional wellness management.

DOI: https://doi.org/10.5281/zenodo.20439509

Business Catalyzer: A Secure Customer Feedback And Analytics Platform For Business Growth

Authors: Karthick Balaji M, Sanjini S P, Varsha S, Gokila M, Dr. M. Rajesh Babu

Abstract: In the contemporary digital economy, businesses of all scales—from informal street vendors to multinational corporations—face the persistent challenge of collecting actionable, continuous, and privacy-preserving feedback from their product users. Conventional review platforms such as Yelp and Google Reviews mandate formal business registration, expose user identities publicly, and provide no structured communication pathway between customer and business. This paper presents Business Catalyzer, role-differentiated feedback management platform engineered using Java Spring Boot, React, MySql, Chart.js. The platform introduces a secure, anonymous, one-on-one case communication channel that enables product users to submit mood-tagged feedback cases against any business entity—registered or otherwise—while ensuring identity privacy and fostering real-time dialogue. A three-graph analytics dashboard comprising the Incoming Ticket Graph, Case Resolution Graph, and Weekly Trend Graph delivers temporally segmented, sentiment-aware performance intelligence to business users. Experimental evaluation on a prototype deployment with 500 synthetic cases across 20 business accounts demonstrates sub-250 ms dashboard refresh latency for datasets up to 10,000 records, a System Usability Scale rating of 4.3 out of 5.0, and 100 percent task completion across five heterogeneous business profiles. The MySQL-backed, stateless architecture supports horizontal scaling to approximately 5,000 concurrent sessions at ten application instances. Business Catalyzer demonstrates that inclusivity, privacy, and data-driven intelligence can be unified within a single, accessible, and scalable feedback infrastructure.

DOI: http://doi.org/10.5281/zenodo.20439586

Time-Series Forecasting of Seasonal Items Sales Using Machine Learning

Authors: Yogeshwaran L, Dinesh Karthick K, Siddarth KM, Vivekanantha Rajeshwaran S

Abstract: Sales forecasting is an essential task in modern business environments, as it helps organizations make informed decisions related to inventory management, production planning, and resource allocation. Accurate prediction of sales becomes more challenging in the case of seasonal products, where demand varies significantly over time due to factors such as weather conditions, festivals, and consumer behaviour patterns. This paper presents a machine learning-based system for forecasting the sales of seasonal items using historical time-series data. The dataset is pre-processed by handling missing values, converting date formats, and organizing data in a structured manner. Feature engineering techniques extract meaningful temporal attributes such as month, day, and year, which play a crucial role in identifying seasonal patterns. A Linear Regression model analyses the relationship between the extracted features and sales values, achieving an R² score of 0.89 with low error metrics (MAE: 10.5, RMSE: 13.4). The system is deployed as a Flask-based web application enabling real-time sales predictions through a user-friendly interface. Results demonstrate that the proposed approach effectively captures seasonal trends and provides reliable predictions suitable for retail, e-commerce, and inventory management applications.

DOI: http://doi.org/10.5281/zenodo.20449169

Personalized Learning Path Recommendation System

Authors: David ray, Shubham Kumar, Vikash Kumar,, Deepjit Ganguly, Sudhanshu Choudhary

Abstract: The rapid expansion of digital learning resources has created both opportunities and challenges for learners. While access to massive open online courses, learning management systems, and skill-based platforms has increased, learners often struggle to identify the most suitable sequence of learning activities aligned with their goals, prior knowledge, and learning preferences. This research paper proposes a Personalized Learning Path Recommendation System (PLPRS) that dynamically recommends learning paths tailored to individual learners. The proposed system integrates collaborative filtering, content-based filtering, and learner profiling techniques to generate adaptive and goal-oriented learning pathways. Experimental evaluation using standard recommender system metrics demonstrates that the proposed approach improves recommendation relevance, learner engagement, and learning efficiency. The study highlights the applicability of personalized learning path systems in higher education and e-learning platforms and outlines future directions for hybrid and AI-driven learning recommendations.

DOI: https://doi.org/10.5281/zenodo.20455927

A Hybrid On-Chainand Off-Chain Architecture For Academic Certificate Validation

Authors: Mr. Ankit Dhakane, Mr. Chetan Maurya, Dr.Jasbir Kaur, Ms. Sandhya Thakkar, Mr. Suraj Kanal

Abstract: The rapid growth of digital education and online credentialing has increased the need for secure and efficient academic certificate verification systems. Traditional verification methods are often centralized, time-consuming, vulnerable to forgery, and dependent on manual institutional processes. Blockchain technology offers a decentralized and tamper-resistant solution for managing academic credentials. This research paper proposes a hybrid on-chain and off-chain architecture for academic certificate validation using blockchain technology and analyzes existing blockchain-based credential verification frameworks such as MIT Blockcerts. The paper examines the architecture, working mechanism, security features, and verification process of Blockcerts, along with its role in preventing certificate fraud and improving transparency. A comparative analysis of traditional and blockchain-based verification systems is also presented to evaluate efficiency, security, and scalability. The findings indicate that blockchain-based academic credential systems enhance trust, reduce verification time, and provide secure ownership of digital certificates. However, challenges such as scalability, interoperability, privacy concerns, and institutional adoption remain significant considerations for large-scale implementation. The study concludes that blockchain technology has strong potential to transform academic certificate management and verification systems in modern educational institutions.

DOI: http://doi.org/10.5281/zenodo.20456095

Autonomous Grievance Redressal System For E-Governance Using AI And Automation

Authors: Diksha Jagtap, Avantika Pachpute, Prof. Shital Y. Mandlik, Dr. Anand A. Khatri

Abstract: Grievances systems for e-governance developed traditionally suffers the issues of duplication of complaints, faulty categorization, delays in resolution, heavy workload of manual intervention and others. The intelligent cloud-based complaint management system powered by artificial intelligence, natural language processing, on-device machine learning, and cloud computing is called AGRS-EG, or autonomous grievance redressal system for e-governance. Through an Android application, users will send complaints containing text, images, and location through GPS. A hybrid classification engine that uses NLP for semantic text understanding and TFLite for on-device image classification achieves a 96.3% combined accuracy. The image validation module rejects irrelevant uploads at 89.5% accuracy. The main contribution is the three-level duplicate detection pipeline category filtering, Haversine-based 50-metre geo-proximity filtering, and AI semantic similarity which results in 93.8% duplicate detection accuracy and 4.1% false duplicate. Any duplicate, based on its severity, number, and type, will cause an escalation in priority based on a dynamic system. AI chatbot solves 87.6% of user queries by itself. Using the Google Maps API the administrative dashboard shows possible hot-spots on the map. The processing was 6.9× faster than manual baseline

DOI: https://doi.org/10.5281/zenodo.20457243

Investigating the Phenomenon of Aircraft Front Wheel Rolling Using Matlab Simulink Software

Authors: Master Van Huy Khuat, Master Trong Son Phan, Master Le Phan

Abstract: Grievances systems for e-governance developed traditionally suffers the issues of duplication of complaints, faulty categorization, delays in resolution, heavy workload of manual intervention and others. The intelligent cloud-based complaint management system powered by artificial intelligence, natural language processing, on-device machine learning, and cloud computing is called AGRS-EG, or autonomous grievance redressal system for e-governance. Through an Android application, users will send complaints containing text, images, and location through GPS. A hybrid classification engine that uses NLP for semantic text understanding and TFLite for on-device image classification achieves a 96.3% combined accuracy. The image validation module rejects irrelevant uploads at 89.5% accuracy. The main contribution is the three-level duplicate detection pipeline category filtering, Haversine-based 50-metre geo-proximity filtering, and AI semantic similarity which results in 93.8% duplicate detection accuracy and 4.1% false duplicate. Any duplicate, based on its severity, number, and type, will cause an escalation in priority based on a dynamic system. AI chatbot solves 87.6% of user queries by itself. Using the Google Maps API the administrative dashboard shows possible hot-spots on the map. The processing was 6.9× faster than manual baseline

DOI: https://doi.org/10.5281/zenodo.20457484

Hybrid Deep Learning For Deepfake Detection: A Systematic Survey

Authors: Komal Khatak, Sonal Beniwal

Abstract: Advancements in generative AI, especially in generative adversarial networks (GANs) and diffusion models, have enabled greater accessibility for users to produce hyper-realistic synthetic media. This has made deep fake detection tools that work with a single modality more vulnerable to adverse real-world environments. To address this, we have implemented a structured survey of hybrid deep learning frameworks that integrate different types of networks, data modalities and representations of the domain. We present a 6-category taxonomy which covers the following areas: (i) CNN architectural hybrids, (ii) CNN-CNN temporal models, (iii) cross-audio-visual modalities, (iv) spatial-frequency hybrid systems, (v) hybrid systems with a forensics perspective, and (vi) adversarial hybrid systems that integrate explainability and adversarial robustness (separation of the model) systems. For the aforementioned areas, we have analyzed the rationale of the design, the fused strategies, and the performance of the systems in relation to the current benchmarks as well as challenges that still persist. Through the analysis of the 45 studies we have examined, we have determined that hybrid models consistently outperform single stream models, especially under compression, domain shifting, and adversarial attacks. Lastly, we have identified challenges that need to be addressed including the generalization gap, the absence of a benchmarking framework, and poor interpretability and we outline systematic and important methods to direct future research with these challenges

DOI: https://doi.org/10.5281/zenodo.20507777

Artificial Intelligence And Copyright Conundrum

Authors: Puneet Sharma, Dr. Alaknanda Rajawat

Abstract: Artificial intelligence is widely acknowledged to be one of the most dramatic technological advancements, which has the potential to be a creator of the world as well as a destroyer. Artificial Intelligence is already infiltrating and having a ripple effect in every strata of society. Things are becoming complex with the use of artificial intelligence by artists for creative works, having afflicting implications for Intellectual Property law. Artificial intelligence can instantly compose music, write blogs, novels, and poetry, and generate paintings and drawings by learning from subsisting work. AI has made it impossible to distinguish between artwork created by humans and one by computer with or without human intervention, raising the question of whether work by AI should be copyright protected; if yes, how? The authorship of computer-generated works was not contested before because the computer program was merely a tool that helped along the creative process; however, with the advent of Artificial intelligence, a computer program is no longer just a tool but stimulates human intelligence to create artistic work by drawing from the already existing database, begging the question of how IPR laws are to be applied to such creative work. This paper assesses the implications of using AI to create works and the potential infringement of existing copyrights. It further attempts to address the issue of the authorship of such AI- generated works and who will have a rightful claim to it. After a comparative analysis of the legal framework of different countries pertaining to AI-generated works, this paper envisages granting authorship to AI in the Indian legal framework.

DOI: http://doi.org/10.5281/zenodo.20487941

LeftOver Link: A Smart Web-Based System For Geo-Tagged Food Redistribution In Local Community

Authors: Aparna Shendkar, Janhvi Dudhane, Neel Dhoble, Diya Bondre, Shreya Dhumal

Abstract: Food wastage and hunger are two major social problems that coexist in today’s world. Large amounts of edible food from households, restaurants, and social events are discarded daily, while many individuals lack access to basic meals. LeftOver Link – Smart Food Redistribution in Locality is a web-based platform designed to bridge this gap by connecting food donors with nearby NGOs or individuals in need. The system enables users to list leftover food by providing details such as food type, quantity, and preparation time. Geo-tagging using Google Maps API allows the platform to identify the donor’s location and notify nearby receivers in real time. Registered receivers can instantly view, claim, and collect available food before it gets spoiled. The platform also includes user authentication, notifications, and a feedback mechanism to ensure transparency and trust among users. By digitizing and automating the food donation process, LeftOver Link minimizes delays caused by manual coordination and improves redistribution efficiency. The proposed system promotes community participation, supports hunger relief efforts, and encourages responsible food management. Overall, the project demonstrates how web technologies can be used effectively to address real-world social challenges and create a sustainable, socially impactful food-sharing ecosystem.

DOI: http://doi.org/10.5281/zenodo.20488182

MDCE-Net: A Multi-Scale Deep Convolutional Ensemble Network for Automated Malaria Parasite Detection and Disease Prediction from Thin Blood Smear Images

Authors: Paramjeet Singh, Dr Raj Kumar

Abstract: Malaria remains one of the most devastating infectious diseases worldwide, claiming over 600,000 lives annually, with the majority of fatalities occurring in sub-Saharan Africa and South Asia. Traditional microscopic examination of blood smears, while being the gold standard for malaria diagnosis is time-consuming, expert-dependent, and prone to human error—particularly in resource-limited settings. This paper proposes MDCE-Net (Multi-scale Deep Convolutional Ensemble Network), a novel deep learning architecture that combines multi-scale feature extraction, Squeeze-and-Excitation attention mechanisms, and ensemble transfer learning using EfficientNetB4 and ResNet50 backbones for automated malaria parasite detection from thin blood smear images. The proposed model was trained and evaluated on a combined dataset of 56,480 cell images sourced from the NIH Malaria Dataset and the Kaggle Malaria Cell Images Dataset, encompassing both parasitized and uninfected cells. Extensive data augmentation strategies including random rotation, horizontal and vertical flipping, zoom, brightness adjustment, and Gaussian noise injection were employed to enhance model robustness. The MDCE-Net architecture achieved a classification accuracy of 99.21%, precision of 99.08%, recall of 99.34%, and F1-score of 99.21% on the test set, outperforming existing state-of-the-art methods including standalone VGG16, ResNet50, and EfficientNetB4 architectures. The model also demonstrates strong generalization performance on independent validation cohorts. This work presents a significant step toward automated, scalable, and deployable malaria diagnostic tools suitable for integration into point-of-care systems in rural and resource-constrained healthcare environments. Source code and model weights are made publicly available to facilitate reproducibility.

Harmfulness of 5G Transmission from Mobile Phone Towers

Authors: Pooja Rani, Vijay Kumar, P. K. Sharma, P. P. Pathak

Abstract: In today’s time the world is going digital. In digital world most people use of mobile phone to convey the information, for which towers receive a signal and transmit it back to mobile set. For this, transmission towers emit the radiation. The increased electromagnetic radiation emitted by towers may possibly affect the human health. Radiation penetrates the human body and is absorbed by the human tissues. To know the possibility of harmfulness we calculate the penetrated field in human body for two different frequencies and for different values of power with the help of SAR formula and the results are compared with the permitted whole-body SAR value recommended by different agencies. Calculating the value of specific absorption rate (SAR) for 5G frequencies from transmission towers it is shown that harmfulness increases in higher frequencies of transmission. Harmfulness reaches to longer distances in higher generation of transmission.

DOI: https://doi.org/10.5281/zenodo.20489380

Synthesis And Characterization Of Iron-Doped Tin(IV) Oxide Nanoparticles: A Systematic Review Of Structural And Optical Properties

Authors: Ashly Mathew

Abstract: Iron-doped tin(IV) oxide (Fe-doped SnO₂) nanoparticles have attracted considerable attention owing to their tunable optical, structural, magnetic, and electronic properties. This review synthesizes findings from recent studies investigating Fe-doped SnO₂ synthesized through sol–gel, hydrothermal, co-precipitation, combustion, and sputtering techniques. The literature demonstrates that Fe incorporation significantly influences crystallite size, oxygen-vacancy concentration, morphology, photoluminescence behavior, and band-gap energy. Most studies report retention of the rutile tetragonal phase with minimal impurity formation. The review highlights the relationship between synthesis conditions and material properties and discusses potential applications in photocatalysis, gas sensing, optoelectronic devices, and energy-storage systems.

DOI: https://doi.org/10.5281/zenodo.20489654

Flutter-Godot Bridge: A Framework For Embedding Godot Engine Games Into Cross-Platform Flutter Applications

Authors: Mr. Yuvraj Singh Sanhotra, Mr. Keval Siddhapura, Dr. Jasbir Kaur, Mr. Suraj Kanal

Abstract: The integration of native game engines into cross-platform mobile application frameworks poses substantial architectural challenges, particularly when reconciling Flutter’s declarative Dart-based UI system with Godot Engine’s native C++ rendering pipeline. Existing embedding approaches suffer from bidirectional communication latency, rendering pipeline conflicts that cause view misalignment, platform-specific imple- mentation complexity, and the inability to host multiple game instances within a single process. This paper presents a novel framework for integrating Godot Engine games into Flutter applications on the Android platform. Unlike prior work that embeds Godot as a PlatformView within the same process, the proposed solution employs a slot-based process isolation archi- tecture where each Godot game instance runs in a dedicated An- droid process. The framework implements bidirectional commu- nication via MethodChannel with a JNI-native C++ bridge, com- prehensive lifecycle management, and performance monitoring through atomic counters. Experimental results demonstrate that process isolation circumvents Godot’s single-instance-per-process limitation, provides crash containment, and enables independent lifecycle management of multiple game sessions while maintaining Flutter host responsiveness. Quantitative evaluation shows that the framework sustains 60fps for simple scenes and achieves sub-5ms round-trip latency for small messages.

DOI: http://doi.org/10.5281/zenodo.20489878

Electric Vehicle Technology

Authors: Professor Deepali Vaidya, Amardip Bandu Raipure

Abstract: Electric Vehicle Technology has emerged as one of the most promising solutions for sustainable transportation in the modern world. Electric Vehicles (EVs) use electric motors powered by rechargeable batteries instead of conventional internal combustion engines that rely on fossil fuels. The increasing concerns regarding environmental pollution, global warming, and depletion of fossil fuel reserves have accelerated the development and adoption of EVs worldwide. Electric vehicles offer numerous advantages, including reduced greenhouse gas emissions, lower operating costs, improved energy efficiency, and decreased dependence on petroleum products. Recent advancements in battery technology, charging infrastructure, and power electronics have significantly enhanced EV performance and reliability. This research paper discusses the concept, history, components, working principles, advantages, disadvantages, applications, challenges, and future prospects of Electric Vehicle Technology.

DOI: http://doi.org/

Food And Nutrition In Hotel Management: A Comprehensive Study Of Quality, Safety, And Customer Satisfaction

Authors: Ms.Suman Rajput, Dr.Sukriti

Abstract: The hospitality industry is undergoing a significant transformation driven by increasing consumer awareness of health, nutrition, and sustainability. Food is no longer evaluated solely on taste and presentation; instead, guests expect nutritionally balanced, hygienically prepared, and ethically sourced meals. This study examines the role of food and nutrition in hotel management, focusing on their influence on customer satisfaction, operational performance, and brand positioning. The research also explores challenges such as cost constraints, supply chain limitations, and lack of staff training, which hinder the effective implementation of nutrition-based practices. Furthermore, it highlights emerging trends including personalized nutrition, sustainable dining, and wellness tourism. Through case studies of leading Indian hotel chains, the study demonstrates how integrating nutrition into hospitality operations enhances competitiveness and long-term sustainability. The findings suggest that nutrition-oriented strategies are essential for modern hotel management and must be incorporated into core operational frameworks.

DOI: http://doi.org/10.5281/zenodo.20504999

AI-Based Early Cancer Screening Using Multi-Modal Data

Authors: Krishna Kumar Amatya, Hemlata, Manish Jung Shah, Aashutosh Prasad Kushwaha, Laxmi Shahi

Abstract: Detecting cancer early greatly improves the chance of survival and decreases the amount of money required to treat it. Most current technologies used in routine cancer screening only use one source of data and therefore have a higher rate of misdiagnosis and longer time until diagnosed. To overcome this obstacle, we propose a new multi-modal artificial intelligence (AI) system that adds multiple sources of data together to form a consolidated platform for diagnosis, such as combining brain MRIs, lung CTs, electronic health records (EHR's), laboratory test results, genomic markers, and clinical notes. Our new system uses convolutional neural networks (CNN's) to classify brain and lung tumours; random forest classifiers to sensor symptom characteristics; and displays results using GradCAM heat maps to provide visual images in a manner that makes sense to the referring physician. Results from experiments conducted with the new system showed a 94.2% accuracy rate (plus or minus), 92.8% sensitivity, 95.1% specificity, and an area under the ROC curve (AUC) of 0.96. The results confirm that multi-modal datasets will improve the methods used for detecting cancer earlier.

DOI: https://doi.org/10.5281/zenodo.20507709

Hybrid Electric Vehicle (HEV)

Authors: Professor Deepali Vaidya, Sujal Bhimrao Gongale

Abstract: A Hybrid Electric Vehicle (HEV) is an advanced automobile that combines an internal combustion engine (ICE) with an electric motor and battery system. The hybrid system improves fuel efficiency, reduces harmful emissions, and enhances vehicle performance. HEVs use regenerative braking technology to recover energy that would otherwise be lost during braking and store it in the battery for future use. With increasing concerns about environmental pollution and depletion of fossil fuels, hybrid electric vehicles have become an important solution for sustainable transportation. This paper discusses the concept, components, working principle, advantages, disadvantages, and future prospects of hybrid electric vehicles.

AI Enabled and Deep Learning-Based Integrated Approach for Early Detection of Breast Cancer

Authors: Dr. Pritesh Patil, Abhiraj Bondre, Gauri Dighe, Kiran Mangde

Abstract: Breast cancer is one of the most common and life-threatening diseases affecting women worldwide, with approximately 2.3 million new diagnoses reported each year. Despite significant improvements in survival rates, early and accurate detection remains a major clinical challenge, especially in resource-constrained healthcare settings where radiologist availability and turnaround times can directly influence patient outcomes. In many hospitals, a single radiologist may be responsible for reviewing hundreds of MRI scans per day under time pressure — a situation that almost inevitably leads to some degree of inconsistency and missed findings. This paper introduces Women Wellness, a fully integrated diagnostic platform designed to tackle these challenges by combining deep learning-based image classification, OpenCV-powered video preprocessing, cloud-hosted parallel inference, and automated clinical report generation within a single deployable system. The platform accepts breast MRI video sequences as input, extracts individual frames on the client side using the HTML5 Canvas API, preprocesses them through OpenCV.js, and routes them through a fine-tuned Convolutional Neural Network hosted as Python Flask microservices. Classification results — spanning normal, benign, and malignant categories — are aggregated using a confidence-weighted voting scheme, and the entire pipeline culminates in a structured, multi-page PDF report generated automatically via jsPDF. In experiments conducted on a curated MRI dataset, the system achieved overall classification accuracy between 92% and 95%, with malignant case sensitivity reaching 96.3%. The complete analysis and report pipeline completes in 45 to 60 seconds per study, compared to roughly 15 to 20 minutes for conventional manual review. Grad-CAM attention maps are embedded directly into each report, enabling radiologists to visually verify which image regions most influenced each classification decision rather than simply taking the model's word for it.

DOI: https://doi.org/10.5281/zenodo.20506325

Ultrasonic Motor

Authors: Professor Deepali Vaidya, Saurabh Vithoba Zade

Abstract: An Ultrasonic Motor (USM) is a special type of electric motor that uses ultrasonic vibrations to produce motion instead of electromagnetic forces. Unlike conventional motors, ultrasonic motors operate through the piezoelectric effect, where certain materials generate mechanical vibrations when subjected to an electric voltage. These vibrations occur at ultrasonic frequencies, typically above 20 kHz, which are beyond the range of human hearing. Ultrasonic motors offer several advantages such as high torque at low speeds, compact size, silent operation, fast response, and the ability to hold position without consuming power. These characteristics make them suitable for precision positioning systems, robotics, medical devices, cameras, aerospace systems, and industrial automation. The technology has gained significant attention due to its ability to overcome limitations of traditional electromagnetic motors. This paper presents the working principle, construction, types, advantages, disadvantages, applications, and future developments of ultrasonic motors. The study highlights the growing importance of ultrasonic motor technology in modern engineering systems requiring high precision and reliability.

DOI: http://doi.org/

A Comprehensive Review on Stroke and It’s Management

Authors: Professor Dr.M.Prasada Rao, Professor Dr.Y.Narasimha Rao, Dr.S.Rajini, M.Vamsi Vardhan

Abstract: Stroke is heterogeneous cerebrovascular disorder characterized by a sudden loss of neurological function resulting from disruption of cerebral blood flow, either due to ischemia(approximately85%of cases) or haemorrhage (about 15%). This comprehensive review focuses on the risk factors that affect the stroke and subtypes for young ischemic stroke patients and their outcomes at the time of discharge. as well as the correlation between these risk factors and stroke subtypes based on the TOAST classification of young strokes. The TOAST classification divides patients with ischemic stroke into five sub groups According to the presume detiological mechanism. The aims of the present study were to evaluate the distribution of the different etiological stroke subtypes in a hospital-based sample of stroke patients, and to investigate the association between important risk factors and stroke subtypes. Stroke prevention strategies is to identify subjects who are at increased risk for stroke and to modify the risk if possible. Although some risk factors are non-modifiable such as Age, gender, family history and race or ethnicity are considered markers for increased stroke risk. Modifiable risk factors for stroke include hypertension, cardiac disease (particularly atrial fibrillation), diabetes, hyper lipidemia, smoking and alcohol consumption. Men have greater stroke incidence than women. This review mainly underlines the need for further research to identify risk for stroke also stroke subtypes and emerging therapies are used to reduce the morbidity and mortality of stroke. Recognition of stroke and its sub types provide the basis for primary, secondary, and tertiary stroke prevention strategies.

Implementation of Artificial Intelligence & CNN for optimization of wall thickness analysis Through Geom-Caliper

Authors: Mohammad Arhum A Haque

Abstract: Wall thickness analysis is a critical step in product design because it directly affects strength, manufacturability, weight, material usage, cooling behaviour, and overall cost. In casting, molding, and similar manufacturing processes, uneven wall thickness can lead to sink marks, warpage, shrinkage, weak sections, and unnecessary material buildup. Traditionally, this check was performed through manual sectioning and visual measurement, making the process slow, repetitive, and highly dependent on the designer’s experience. Geom-Caliper improved this workflow by enabling automated wall thickness inspection directly on three-dimensional computer-aided design models, allowing faster and more repeatable identification of thin and thick regions. However, optimization of wall thickness still depends largely on manual interpretation, engineering judgment, and repeated trial-and-error modifications. This research proposes an artificial intelligence-based framework to optimize wall thickness analysis through Geom-Caliper. The objective is not to replace Geom-Caliper, but to extend its capability by combining accurate geometric thickness measurement with intelligent prediction. Geom-Caliper is used to generate reference thickness data from CAD models, while artificial intelligence is trained to learn patterns from past models and identify regions likely to become thin, thick, or manufacturability critical. For this purpose, the three-dimensional geometry is converted into voxel-based data so that a three-dimensional convolutional neural network can learn spatial relationships between shape features and wall thickness behaviour. The proposed framework supports early detection of problematic zones before final design validation, reducing manual review effort, improving design consistency, and supporting efficient material usage for advanced manufacturability engineering.

DOI: http://doi.org/

Data Structures in Artificial Intelligence: Foundations, Applications, and Future Directions

Authors: Pankaj Diwakar Katre

Abstract: Artificial Intelligence (AI) has become one of the most transformative technologies of the modern era. Now a days it is the most wide field to explore. The effectiveness of AI systems depends not only on algorithms and computational power but also on the underlying data structures that organize, store, and process information efficiently. Data structures provide the foundation for machine learning, deep learning, natural language processing, robotics, and intelligent decision-making systems. This paper explores the significance of data structures in AI, examining commonly used structures such as arrays, linked lists, stacks, queues, trees, graphs, hash tables, and heaps. The study discusses their applications in AI models, search algorithms, knowledge representation, and optimization techniques. Furthermore, recent developments in large language models and graph-based AI systems are analyzed to demonstrate the evolving role of advanced data structures in intelligent computing. The paper concludes by highlighting future research opportunities in scalable and adaptive data structures for next-generation AI systems.

DOI: https://doi.org/10.5281/zenodo.20509362

Smart Grid Technology

Authors: Professor Deepali Vaidya, Pranav Pradeep Jinde

Abstract: The Smart Grid is an advanced electrical power system that integrates modern communication, automation, and information technologies with the traditional power grid. It enables efficient generation, transmission, distribution, and consumption of electrical energy while improving reliability, security, and sustainability. Smart Grid technology uses sensors, smart meters, automated controls, and communication networks to monitor and manage power flow in real time. The increasing demand for electricity, integration of renewable energy sources, and the need for energy conservation have accelerated the development of Smart Grids worldwide. Smart Grid systems help reduce power losses, improve fault detection, enhance energy efficiency, and support the integration of solar and wind energy. This technology also empowers consumers by providing detailed information about electricity usage, enabling better energy management. Despite challenges such as cybersecurity risks and high implementation costs, Smart Grids represent the future of modern power systems. This paper discusses the concept, components, working principle, advantages, applications, challenges, and future scope of Smart Grid Technology.

DOI: https://doi.org/10.5281/zenodo.20522480

IOT-X Based Intelli City Management System

Authors: Professor Mahesh Dumbhere, Mr. Shubham Nanne, Mr. Swayam Bommewar, Ms. Pooja Prajapati, Mr. Ritik Khairkar, Mr. Mohan Lavhale

Abstract: Smart city management has become one of the most important areas of research due to increasing urban population, infrastructure challenges, and the growing demand for efficient public services. Traditional complaint management systems used in municipalities are often slow, manual, and lack transparency, causing delays in solving public issues such as street light failures, garbage collection problems, water supply issues, drainage blockages, and road damage. The proposed project, “IOT-X Intelli City Management System,” introduces a smart web-based platform designed to improve complaint management and city administration through automation, analytics, and real-time monitoring. The system allows citizens to register complaints online with location details, images, and issue categories. Based on the issue type, the system automatically assigns the complaint to the appropriate department such as Electrical, Water, Sanitation, Drainage, Traffic, Transport, Garden, or PWD. The platform includes intelligent duplicate complaint detection, complaint prioritization, delayed complaint monitoring, notification updates, PDF/Excel report generation, and analytics dashboards. The project is developed using Flask, MySQL, HTML, CSS, JavaScript, and Chart.js. The admin dashboard provides department-wise analytics, complaint trends, delay analysis, and monthly reports. Department officers can manage assigned complaints, while users can track complaint status in real time. The system improves transparency, reduces complaint resolution time, enhances departmental coordination, and supports data- driven smart city governance. Future enhancements may include IoT sensor integration, GPS tracking, AI-based prediction systems, mobile applications, and SMS/email notification services.

DOI: http://doi.org/

Automatic Number Plate Detection Using Yolov8

Authors: Tanuj Kumar Kusmi, Manish Kumar, Vijay Kumar Gautam, Martand Mishra, Abishek Kumar Singh, Satyam Kumar

Abstract: The traffic within cities has become worse. So we should implement automatic number plate recognition (ANPR). It can be used for traffic management, toll collection, police help and smarter. It has been proved that, some traditional image processing techniques and some of the deep learning approaches have not shown success due to their limitations like lighting condition, blocked license plates, and skewed license plates. This project proposes a system that utilizes YOLOv8. This system uses YOLOv8 and character recognition to detect license plates in the images and videos. So, it acts as an "eye" for identifying license plates. Automatic Number Plate Detection system, has demonstrated success in locating license plates, and it also performs efficiently under various lighting conditions including night, even under blocked plates. The combination of YOLOv8 and character recognition techniques offers a real-time solution for detection and identification of license plates in videos and images for effective traffic management and police assistance for instance tracing vehicles through[27]. The system has been designed to work for global number plates and can prove to be an effective solution for license plate detection with its fast speed technology. Good for Parking systems, toll collection and traffic management. Managing city traffic is a critical issue today, an effective Automatic Number Plate Detection system like the one using YOLOv8 and number plate could provide an efficient solution for this issue using number plates. Parking systems and toll collection can really benefit from this. Managing the traffic in a city is a problem nowadays. A good Automatic Number Plate Detection system that uses YOLOv8 and number plates can help solve this problem. work well with. To make the model work better some techniques have been used. These include flipping the image making it bigger or smaller and changing the brightness. This helps the model to work in different situations. The system can detect number plates fast and it is very accurate. It can detect number plates with an accuracy of 96.4%. It can process over 60 frames per second on a standard computer. This is very good, for Automatic Number Plate Detection systems that use number plates. The YOLOv8-based ANPR system does a job on benchmark datasets and real-world test scenarios. It is better than systems that use CNN or earlier YOLO versions when it comes to how fast it detects things how accurate it is and how well it works in different situations. The YOLOv8-based ANPR system can handle license plates that're at an angle partly covered or different sizes and it works well in various lighting conditions[16].

DOI: http://doi.org/10.5281/zenodo.20524663

Student Management System with Artiffical Intelligences

Authors: Assistant Professor Vivek Kumar, Pratham Bansal, Shraddha Rastogi, Nida Khan, Simran Michael

Abstract: The rapid digital transformation in the education sector has created a growing need for intelligent systems that can efficiently manage student information while supporting academic success. Traditional Student Management Systems (SMS) primarily focus on maintaining records such as attendance, examination results, course enrollment, and student profiles. However, these systems often lack the capability to analyze data and provide meaningful insights for decision-making. To address this limitation, this research proposes an AI-Based Student Management System that integrates Artificial Intelligence (AI), Machine Learning (ML), and Data Analytics to improve educational administration and student performance monitoring. The proposed system is designed to automate routine administrative tasks, reduce human errors, and enhance the overall efficiency of educational institutions. It collects and processes data related to student attendance, academic performance, assignment submissions, and classroom activities. Using machine learning algorithms, the system can predict student performance, identify students who may be at academic risk, and generate personalized recommendations for improvement. The system also includes an AI-powered chatbot that provides instant responses to student queries regarding courses, schedules, attendance, and academic progress. In addition, the system offers advanced analytical dashboards and reporting tools that assist teachers, administrators, and parents in monitoring student development. Real-time notifications and alerts help ensure timely intervention when students show signs of poor performance or irregular attendance. By leveraging predictive analytics, the proposed solution enables institutions to make data-driven decisions and improve educational outcomes. The implementation of an AI-Based Student Management System contributes to creating a smart educational environment that promotes personalized learning, effective resource management, and proactive student support. Although challenges such as data privacy, security, and implementation costs exist, the benefits of improved efficiency, accuracy, and student engagement make this approach highly valuable. The proposed system represents a significant step toward the future of intelligent educational management and digital transformation in academic institutions.

DOI: http://doi.org/

Generative AI Meets Cross-Brand Fit Intelligence: A User-Centric Framework for Outfit Recommendation with Occasion and Weather Awareness

Authors: Pranjal Nilesh Belalekar, Om Arun Yadav, Dr. Jasbir Kaur, Assistant Professor Suraj Kanal

Abstract: Online fashion retail suffers from inconsistent sizing across brands and difficulty in composing context-appropriate outfits. Existing recommender systems address either fit or style, but not both in an integrated manner. We present FitGen, a generative AI framework that combines cross-brand size intelli-gence with occasion- and weather-aware outfit recommendation. FitGen collects user body measurements and style preferences, maps them to brand-specific sizes using a weighted Euclidean distance heuristic, and generates personalized outfit descriptions via GPT-4o-mini and corresponding visualizations via DALL-E-3. All interactions occur within a privacy-first Streamlit dashboard. A controlled user study (N=100) demonstrates that FitGen achieves 78.3% fit accuracy across five brands, with precision and recall values of 0.79 and 0.78 respectively, a 4.6/5 user satisfaction score, and a 92% reduction in self-reported size anxiety. The end-to-end latency is 3.2 seconds. We compare our results with existing commercial solutions and academic models, highlighting the trade-offs between transparency and accuracy. While limitations exist, including synthetic measurement distributions, a heuristic size mapper, and the absence of live e-commerce APIs, the results indicate that combining generative AI with explicit fit modeling significantly enhances the online fashion shopping experience. The prototype, source code, and a demo video are made publicly available to facilitate further research.

DOI: https://doi.org/10.5281/zenodo.20529938

Trust Based Routing Security Against Sybil Attacker In FANET

Authors: Avinash Singh, Dr. Ankur Pandey

Abstract: FANET UAVs temporarily connect to other UAVs to transfer data. UAVs have limited memory and capability. In dynamic networks, UAV routing and packet transfer pose security issues. UAVs are subject to assaults and have variable speeds. FANET attacks vary. The malicious nodes isolation or Syble attacker is a packet-dropping assault that steals other nodes' unique identity (ID) and exploits it for network attacks. This paper provided two modules: attacker confirmation and detection and prevention. FANET's first Route and Data Security for Hostile Behaviour Prevention (RDSBP) method for Sybil attacker. Implement a base station device behavior analysis route security mechanism to detect this type of malicious conduct. The attacker node alters the routing table to choose a bogus path and diverts the route by altering the routing packet's next hop address. The main purpose of RDSPB is to identify the attacker. Comparing RDSBP to LOAD and LPAR, innovative security solution performs better. The attacker detection and prevention approach calculates UAV trust based on network function in the second module. The Trust & Energy Aware Secure Routing (TEASR) solution isolates FANET Sybil malicious nodes. Due of its limited energy, UAVs cannot replace batteries instantly. Attackers created retransmission potential, affecting energy. The attacker targets many real UAVs. Novel TEASR outperforms secure and trusted AODV. Secure AODV outperforms Trusted AODV and pure AODV routing. RDSBP and TEASR have the highest PDR and use less energy while routing.

DOI: http://doi.org/10.5281/zenodo.20536728

Automatic Floor Cleaning Robot Using ESP32

Authors: Dr. M.A. Natu, Abhinandan Kumbhar, Sarvesh Vikas Gathe, Dinsha Kazi, Shubham Mohite

Abstract: This paper presents the design and development of a cost-effective autonomous floor cleaning robot based on the ESP32 microcontroller. The system integrates obstacle detection, navigation, vacuum cleaning, and mopping functionalities into a single compact unit. Ultrasonic sensors enable real-time obstacle avoidance, while a grid-based navigation algorithm ensures efficient cleaning coverage. The robot can operate in both manual and autonomous modes using a mobile application interface developed on the Blynk IoT platform. Experimental results demonstrate effective performance in indoor environments with minimal human intervention. The proposed system offers an economical alternative to commercially available robotic cleaners.

DOI: https://doi.org/10.5281/zenodo.20537060

Smart Home Automation

Authors: Professor Deepali Vaidya, Ujwal Ramesh Ratnaparkhi

Abstract: Smart Home Automation is a modern technology that allows homeowners to control and monitor household appliances and systems automatically or remotely through smartphones, computers, or voice assistants. The system integrates various devices such as lighting, fans, air conditioners, security cameras, door locks, and sensors into a centralized network. The main objective of smart home automation is to improve convenience, security, energy efficiency, and comfort. Internet of Things (IoT) technology plays a significant role in connecting devices and enabling communication between them. Users can monitor and control home appliances from any location through internet connectivity. Smart home automation systems can also detect unusual events such as gas leakage, fire, or unauthorized entry and immediately alert homeowners. With the advancement of wireless communication technologies like Wi-Fi, Bluetooth, ZigBee, and cloud computing, smart homes are becoming more affordable and accessible. This research paper discusses the architecture, components, working principles, applications, advantages, challenges, and future scope of smart home automation systems.

DOI: http://doi.org/

Iot Enabled Smart Waste Monitering System

Authors: Prof. Anand Donald, Mr. Lahu Ajmera, Mr. Rohit Khandarkar, Ms. Shruti Aitlawar, Ms. Ekta Nikure, Ms. Payal Chide

Abstract: In today’s rapidly growing urban environment, waste management has become one of the major challenges faced by municipalities and smart cities. Traditional waste collection methods are inefficient because garbage bins are checked manually, which often results in overflowing bins, foul smell, environmental pollution, and unhealthy surroundings. To solve this problem, this project proposes an IoT Based Smart Waste Monitoring System, an intelligent waste management solution that uses Internet of Things (IoT) technology for real-time monitoring of garbage bins. The system uses ultrasonic sensors to detect the garbage level inside the dustbin and a Node MCU/ESP8266 microcontroller to process and transmit the data through Wi-Fi connectivity. When the garbage level exceeds the predefined limit, the system automatically sends notifications to municipal authorities through a mobile application or web dashboard. The system also supports cloud monitoring for efficient waste collection and route planning. This solution improves cleanliness, reduces manual monitoring, minimizes fuel consumption, saves operational cost, and supports smart city development. The proposed system is low-cost, efficient, reliable, and easy to implement using IoT technologies, sensors, and cloud communication.

Design of an Experimental Device for Evaluating the Performance Characteristics of an Aircraft Propeller Propulsion System

Authors: Truong Thanh Nguyen, Trong Son Phan

Abstract: Aircraft propeller propulsion systems are widely used in aviation and unmanned aerial vehicles. However, practical training and research on variable-pitch propellers remain limited due to the complexity and cost of full-scale systems. This paper presents the design of a laboratory-scale experimental platform for investigating aircraft propeller propulsion performance. The system integrates a BLDC motor, variable-pitch propeller mechanism, servo pitch control, thrust and torque sensors, rotational speed measurement, electrical power monitoring, and real-time data acquisition. The platform enables evaluation of thrust, power consumption, efficiency, aerodynamic loading, and reverse-thrust characteristics under various operating conditions.

DOI: http://doi.org/10.5281/zenodo.20537512

Farm App: Empowering Farmers with Technology

Authors: Jayant Dane, kiran pustode

Abstract: Agriculture is one of the most important sectors of the economy, supporting food production and rural livelihoods. However, farmers face several challenges such as climate change, low productivity, lack of market access, water scarcity, and insufficient knowledge about modern farming techniques.Technology has emerged as a powerful tool to address these challenges and improve agricultural efficiency. This research paper explores how modern technologies such as Artificial Intelligence (AI), Internet of Things (IoT), drones, mobile applications, precision farming, and digital marketplaces are empowering farmers. The study examines the benefits, challenges, and future opportunities of technological adoption in agriculture. The findings suggest that technology increases productivity, reduces costs, improves decision-making, and enhances farmers’ income and sustainability. The paper concludes that wider technological awareness, government support, and digital literacy are essential for transforming agriculture and empowering farmers globally.

Eternal Voice: An Adaptive Multi-Modal Framework For Personalized Emotional Memory Preservation

Authors: Suyash Arvind Lothe, Bhavya Ketan Doshi, Dr. Jasbir Kaur, Sandhya Thakkar, Suraj Kanal

Abstract: The preservation of human identity through gen-erative AI presents unique challenges in maintaining vocal fidelity, emotional nuance, and ethical integrity. While individual technologies for text-to-speech (TTS) and large language models (LLMs) are mature, their integration into a cohesive, user-centric framework for digital legacy remains under-explored. This paper presents Eternal Voice, an adaptive multi-modal framework that tightly couples speaker-adaptive voice synthesis with an emotion-aware conversational agent. The key contribution of this work is a novel integration pipeline. It utilizes a fine-tuned Llama-3-8B model with emotion vector conditioning and a Tacotron 2/WaveNet vocoder stack optimized for low-resource speaker adaptation (SV2TTS). We provide a comprehensive technical evaluation, including an ablation study that quantifies the impact of emotion injection on response naturalness. Experimental results on the LibriTTS and ESD datasets demonstrate a Mean Opinion Score (MOS) of 4.6 ± 0.2 for voice similarity. Moreover, the dialogue coherence improves by 18% over standard LLM baselines. Critically, this paper includes a rigorous discussion of failure modes, deepfake countermeasures, and the limitations of current emotional AI in handling ambiguous human affect.

DOI: http://doi.org/10.5281/zenodo.20541580

Wireless Charging System For Electric Vehicles: Technologies, Challenges, And Future Directions

Authors: Mini, Dr. Charu, Garvit sharma

Abstract: Wireless charging systems for electric vehicles (EVs) are gaining significant attention as a transformative technology that enhances user convenience, safety, and automation. Unlike conventional conductive charging systems, wireless power transfer (WPT) eliminates the need for physical connectors, thereby reducing maintenance issues and enabling seamless energy transfer. This paper provides an in-depth analysis of various wireless charging techniques such as inductive power transfer, resonant inductive coupling, and capacitive power transfer. It also discusses system architecture, control strategies, efficiency optimization, misalignment issues, and electromagnetic compatibility. Furthermore, the integration of wireless charging with smart grid technologies and renewable energy sources is explored. The paper concludes with future research directions aimed at improving efficiency, scalability, and commercialization.

DOI: http://doi.org/10.5281/zenodo.20551863

6G Technology And Future Wireless Communication Systems

Authors: Prof. K. M. Jadhav, Ms. Dhanashri Kulkarni, Ms. Shreya Jadhav, Ms. Sneha Jagtap, Mr. Om Mali, Mr. Akshay Kumbhar, Mr. Utkarsh Kumbhar, Mr. Pratik Karande, Mr. Aryan kamble, Mr. Satish Lokhande

Abstract: The sixth generation (6G) of wireless communication technology represents a transformative leap beyond the capabilities of current 5G networks, promising data rates of up to 1 Terabit per second (Tbps), sub-millisecond latency, and seamless integration of artificial intelligence (AI), terahertz (THz) communications, intelligent reflecting surfaces (IRS), satellite networks, and Internet of Things (IoT) ecosystems. This paper explores the technological foundations, enabling paradigms, key challenges, and far-reaching applications of 6G communication systems. A qualitative and analytical research approach is adopted, drawing from recent IEEE publications, ITU reports, and global standardization initiatives. The study identifies significant research gaps including immature THz propagation models, insufficient AI-native architectures, security vulnerabilities in post-quantum environments, and deployment feasibility constraints in developing regions. Through systematic synthesis of the literature, this paper provides a comprehensive reference for researchers, engineers, and policymakers navigating the 6G landscape.

DOI: http://doi.org/10.5281/zenodo.20552111

Health Monitoring Dashboard: Data Analytics & Visualization — Architecture, Implementation, and Future Directions

Authors: Dr Raj Kumar, Aarchi, Karuna Rajput, Shagun Kamboj

Abstract: The rapid rise in digital health records has sparked a strong need for smart tools that convert unprocessed body data into practical medical decisions. This study introduces the Health Monitoring Dashboard (HMD), a unified web app combining data analysis, visual interaction, and machine learning predictions. Developed using Python (v3.x), the platform relies on Streamlet for real-time responses, Plotly Express for rich charts, Pandas for efficient data handling, and Joblib to manage trained models. The interface features six components: filtering user conditions, aggregating performance metrics, displaying time-based trends and cross-data relationships, applying rule-driven health rules, and offering a machine learning testing environment. Tests confirm the system accurately handles inputs such as heart rate, blood pressure, sleep habits, steps taken, and physical activity with live updates achieved in less than a second on typical personal computers. Future improvements include deeper ML model support through Joblib workflows, connecting directly to fitness devices like Fitbit or Garmin, deploying securely on cloud services with backend databases, and enabling multi-user access based on roles. The results show that accessible Python tools can build reliable healthcare analytics systems and offer a replicable blueprint for others

DOI: http://doi.org/

A Review Paper On Space Traffic Management

Authors: Prof.K.M. Jadhav, Mr.Siddhesh Masal, Ms.Madhura Patil, Mr.Pranav Pawar, Ms.Pratiksha Pawar, Ms.Sanchita Patil, Ms.Vaishnavi Mulik, Mr.Mayur Mane, Mr.Samarth Patil, Mr.Saurabh Rathod

Abstract: Space has always fascinated humanity, driving curiosity, innovation, and technological advancement. From early satellite launches to modern space missions, human exploration beyond Earth has expanded rapidly. However, this progress has introduced a serious and often overlooked problem—space debris. As more satellites, rockets, and missions are launched, non-functional objects and fragments accumulate in orbit, especially in Low Earth Orbit (LEO), posing significant risks to operational spacecraft and future exploration. Space Traffic Management (STM) has been developed to monitor and regulate space activities, aiming to reduce collisions and manage orbital congestion. Despite these efforts, the increasing rate of technological development and frequent space missions have made existing systems insufficient. However, This growth has also created a serious issue called debris, threatens not only space infrastructure but also the long-term sustainability of space operations. This project focuses on analyzing the causes and impact of space debris in Earth’s orbit. It examines current management systems and identifies their limitations. Furthermore, it explores effective methods and strategies to control and reduce space debris, ensuring safer and more efficient use of space for future generations.

DOI: http://doi.org/10.5281/zenodo.20555586

CFD Investigation Of Pressure Drop, Friction Factor, And Thermal–Hydraulic Performance Of Geometrically Modified Twisted Tape Inserts

Authors: Ajay Malviya, Dr. Satnam Singh

Abstract: This CFD investigation examines the pressure drop, friction factor, and thermal–hydraulic performance of geometrically modified twisted tape inserts in a circular pipe. Twisted tape inserts are passive heat transfer devices that improve fluid mixing by producing swirl flow, turbulence, and stronger wall-fluid contact. However, such modifications may also increase pressure drop and friction factor, which can raise pumping power. Therefore, this work aims to identify a twisted tape design that provides better heat transfer with acceptable flow resistance. Four geometries were studied: Plain Twisted Tape, Double-Hole Perforated Twisted Tape, Curved-Slot Twisted Tape, and Multi-Hole Perforated Twisted Tape. A 400 mm long pipe with 42 mm inner diameter and 44 mm outer diameter was modeled with a full-length twisted tape insert. The mesh was generated in ANSYS Fluent Meshing with 472,350 cells. Water was used as the working fluid, while the pipe and twisted tape were assigned as aluminum. The standard k-epsilon turbulence model was used for turbulent flow simulation. The Multi-Hole Perforated Twisted Tape showed the highest heat transfer rate of 57.67 W and the best thermal–hydraulic performance value of 1.238. This design is the most suitable among the tested geometries.

DOI: http://doi.org/10.5281/zenodo.20556899

AI-Enabled Smart Wearable System for Continuous Monitoring of Cardiac Patients

Authors: Research Scholar Anitha Udayakumar, Professor Senthil Kumar Thillaigovindhan

Abstract: Despite significant advances in health care systems and technology, cardiovascular diseases (CVDs) continue to be the most common causes of death globally, with over 17.9 million fatalities per year. Continuous monitoring of cardiovascular health is vital for early identification of abnormalities but is limited by bulky, non-continuous nature of existing solutions such as the Holter monitor. In this work, we present a new intelligent wearable that provides continuous, real-time monitoring of the heart activity using an edge-AI approach based on an ECG, photoplethysmography (PPG), and accelerometer sensors. Specifically, our solution incorporates an energy-efficient multi-sensor wearable that transmits data to an edge-AI processor, running a lightweight 1D-CNN-LSTM model, for real-time classification of cardiac arrhythmia. The presented federated learning technique allows personalizing models across the population to guarantee individual-level privacy while maintaining high performance. In extensive experiments conducted using MIT-BIH Arrhythmia Database (n=49 patients) and a clinical trial (n=120 patients), our solution demonstrated up to 98.3% sensitivity and 97.6% specificity in recognizing 11 classes of cardiac arrhythmias, as well as end-to-end latency less than 100ms.

DOI: https://doi.org/10.5281/zenodo.20557672

Microbial Intelligence and Communication in Hidden Ecosystem

Authors: Assistant Professor Dr. J. Karthegaa,, Assistant Professor Neema Gopal

Abstract: Underneath the surface of obvious ecosystems is a whole other world of amazing complexity and intelligence—the microbial world. Microbes, which include bacteria, fungi, protists, and archaea, are not lone individuals but rather intricate communication networks, which control ecosystem behavior, biogeochemical cycling, and even their hosts' well-being. In this paper, we review existing knowledge about microbial intelligence and communication systems, especially those related to quorum sensing (QS), biofilm formation, electrochemical signaling, and cross-kingdom interactions. We propose a computational approach to modeling microbial communication networks based on graph theory and agent-based modeling (ABM), allowing us to model information exchange and collective decision-making in the hidden ecosystems of soil microbiomes, deep sea hydrothermal vents, and human gut microbiota. Using metatranscriptomics and chemical signal profiling, we illustrate that microbial populations are capable of exhibiting intelligent behavior in terms of problem solving, memory, and optimization, akin to neural networks.

DOI: https://doi.org/10.5281/zenodo.20557745

Deep Learning-Based Bioinformatics Framework for Early Disease Prediction Using Genomic Data

Authors: Assistant Professor Pushpa Rajita G, Research Scholar Anitha Udayakumar

Abstract: With the introduction of high throughput genomics sequencing technologies, there has been an explosion of genetic information. However, predicting any useful clinical insights from this wealth of information poses major challenges. Prediction of diseases at an early stage using genomics analysis would allow us to take preventive actions prior to manifestation. In this paper, we propose a complete deep learning-based bioinformatics framework for disease prediction using DNA sequencing data. This framework uses three key elements including (1) a hybrid CNN-RNN architecture for performing variant calls as well as extracting important features from the sequencing information, (2) a Graph Attention Network (GAT) for modeling interaction networks between genes, and (3) a multi-modal fusion layer combining genomic, epigenomic, and clinical information. Tested on three large scale databases (TCGA for cancers, UK Biobank for cardiovascular disease, and ADNI for Alzheimer's disease), our proposed framework obtained AUC values of 0.956, 0.934, and 0.921 respectively, which is much higher than traditional GWAS approach as well as several deep learning baselines. Additionally, we have proposed novel attention maps providing biological insights of pathogenic variants and their interaction network. We have conducted future proofing of this model on 500 patients at high risk.

DOI: http://doi.org/

Qr Code Generator for Medical Emergency

Authors: Prof. Tejas Moon, Ms. Sakshi Kankuntla, Ms. Manswi Meshram, Ms. Yasmeen Pathan, Ms. Supriya maidam, Ms. Sumita Patil, Ms. Kashish Gorghate

Abstract: In today’s fast-paced world, medical emergencies require immediate action and rapid access to patient medical history. In many emergency situations, patients may be unconscious or unable to communicate important medical details such as blood group, allergies, medications, or emergency contacts. This delay in obtaining patient information can result in incorrect treatment, delayed care, and even loss of life. To solve this problem, this project proposes a QR Code Generator for Medical Emergency; a smart web-based application designed to store and provides instant access to critical patient medical data through QR technology. The system allows users to enter personal and medical information such as full name, blood group, allergies, medical notes, emergency contact details, patient photo, and medical report links. Once entered, the system generates a unique QR code that securely stores this information. During an emergency, healthcare professionals or first responders can scan the QR code using any smartphone or QR scanner and instantly retrieve the patient’s medical information. The system also provides a downloadable PDF medical report for hospital documentation and future reference. This solution improves healthcare accessibility, minimizes response time, enhances patient safety, reduces paperwork, and provides a low-cost digital healthcare solution using HTML, CSS, and JavaScript, QRCode.js, jsPDF, and web technologies.

AI – Powered Personalization Care Recommendation Engine

Authors: Somit Kumar Yadav, Rohan Choudhary, Sri Krishna Mishra, Minku Kumar, Subhash Kumar Yadav, Himanshu Kumar, Rahul Kumar shah

Abstract: There are a wide range of variables in the field of health care considered by this AI system such as age, gender, height, weight, body mass index, hours of sleep, hydration techniques, stress levels, physical activities, health symptoms, previous diseases to identify potential healthcare risks that can occur in the future and give personalized recommendations on how to take proper care of your health. Predictive healthcare analysis techniques implemented in this project consist of machine learning algorithms such as Logistic Regression, Decision Tree, Random Forest, and XGBoost for predictive analysis and comparison [7], [8], [24]. Predictive analysis is combined with healthcare rules to establish adaptive health care guidelines related to such areas as food management, performing physical exercises, changing hydration practices, decreasing stress and implementing preventive health care strategies. Such technologies as React (frontend framework), Flask (backend APIs), MySQL (healthcare data management) and Scikit-Learn (implementing machine learning models) are implemented to design this project. Methods such as using JSON Web Token-based authentication help with secure healthcare data processing [20], [27]. This AI system is not intended to be used as diagnostic or treatment tool but as a decision-making health care support system to improve preventive health care and personalize it using AI [2], [12].

DOI: http://doi.org/10.5281/zenodo.20560226

Student Management System: A Web-Based Solution for Digital Academic Administration

Authors: Dr Raj Kumar, Ashutosh Pandey, Arpit Chauhan, Bhanu Pratap Yadav, Divyanshu Raj

Abstract: The Student Management System (SMS) is a web-based application designed to digitise and streamline the administrative operations of educational institutions. Traditional methods of managing student records—relying on registers, paper-based attendance sheets, and manual report generation—are time-consuming, error-prone, and difficult to scale. This paper presents the design, development, and evaluation of a web-based SMS built using HTML5, CSS3, JavaScript, PHP, and MySQL, hosted on an XAMPP local server environment. The system provides five core modules: student record management, attendance management, marks and result management, authentication and security, and notification and report generation. Role-based access control is implemented for three user types—administrators, teachers, and students. The system follows an Agile development methodology with iterative sprint cycles. Results demonstrate that the SMS significantly reduces manual workload, improves data accuracy, and provides a centralised, secure platform for academic administration. The paper further discusses limitations, such as dependence on stable internet connectivity and basic computer literacy requirements, and proposes future enhancements including mobile application support, biometric attendance, AI-based performance tracking, and cloud database integration.

Road Cycling Race

Authors: Akkenapalli Maheshwari, Gaddam Harshith Reddy, Gunda Sathya Vamshi

Abstract: This project focuses on the analysis and visualization of a Road Cycling Race dataset using Tableau. The dataset contains information about race winners, teams, nationalities, average speeds, distances, stages, entrants, and finishers. Various visualization techniques such as line charts, bar charts, box plots, maps, funnel charts, and timeline charts were used to explore patterns and trends in the data. The analysis helped identify performance variations, winner distributions, and historical race.

A Comprehensive Review And Update On Crohns Disease

Authors: Dr. Manchineni Prasada Rao, Dr. V. Rajini, Dr. Y. Narasimha Rao, Shaik Mohammad Yasin

Abstract: In the United States, it is currently estimated that about 1.5 million people suffer from Inflammatory Bowel Disease, causing considerable suffering, mortality and economic loss every year. Yet the cause of IBD is unknown, and until we understand more, prevention or cure will not be possible. There is a lot of variation in the incidence and prevalence of Crohn’s Disease based on geographic region, environment, immigrant population, and ethnic groups. The annual incidence of Crohn’s Diseases in North America is reported to be 3.1–20.2 per 100,000 with a prevalence of 201 per 100,000 population. Based on the epidemiological, genetic and immunological data, Crohn’s Disease is considered to be a heterogeneous disorder with multifactorial etiology in which genetics and environment interact to manifest the disease. Several genes have been studied so for with respect to Crohn’s Disease, but thus far the strong and replicated associations have been identified with NOD2, IL23R and ATG16L1 genes. The risk factors implicated with Crohn’s Disease include smoking, low fiber- high carbohydrate diet, altered microbiome and medications such as non-steroidal anti-inflammatory drugs.

Voltage Stabilization Of A DC Microgrid Feeding An EV Charging Station Using A Systematically Tuned Fractional-Order Controller

Authors: Sandeep Khanduri, Dr. Himanshu Maithani, Dr Charu

Abstract: High-power electric vehicle (EV) fast charging causes problems like CPL behavior, which destabilizes DC microgrids because of negative incremental impedance. In this paper, a control scheme for DC bus voltage stabilization in an EV integrated DC microgrid based on a fractional-order proportional integral derivative (FOPID) control approach has been proposed. First, a nonlinear averaged modeling considering boost converter dynamics and the characteristic of CPL is performed. Then, linearized small signal models are used to investigate the limits of stability along with critical parameters. A multi-objective tuning approach is then employed to tune a PI, PID, and FOPID controller under the same operation conditions. The performance comparison has been done in the time domain, frequency domain, sensitivity analysis, and nonlinear disturbance analysis. It is concluded that FOPID controller performs better than PID and PI in terms of phase margin, disturbance rejection, and robustness, while providing a good compromise in other parameters.

DOI: http://doi.org/10.5281/zenodo.20569013

Big Data Analytics In E- Commerce

Authors: Vaishnvee Prasad Pawar, Pravin J. Chougule

Abstract: BDA has emerged as a transformative technology in the e-commerce industry, enabling organizations to process and analyze vast amounts of customer and operational data.The expansion of e-commerce platforms has led to a substantial increase in the volume of information generated through online purchases, website visits, product searches, customer feedback, social networking activities, and electronic payment transactions. . This study examines the impact of BDA on e-commerce operations in India between 2018 and 2025.Using secondary data collected from government reports, industry publications, academic journals, and e-commerce case studies, the research evaluates how analytics contributes to customer personalization, demand forecasting, inventory optimization, fraud detection, and strategic decision-making. The findings indicate that organizations utilizing analytics-driven approaches achieve higher customer satisfaction, improved operational efficiency, and increased revenue generation.The study also identifies key challenges including data privacy concerns, infrastructure costs, cybersecurity threats, and shortages of skilled professionals. The research concludes that BDA has become a critical success factor for modern e-commerce businesses and will continue to influence digital commerce through integration with Artificial Intelligence, Machine Learning, and cloud-based technologies.

Simulation And Implementation Of Hybrid Error Correction Codes For Space Applications

Authors: Shaik Jani Begum, J Ravisankar

Abstract: This research aims to develop a new EDAC system that can detect and repair mistakes in space-grade onchip memory produced by cosmic radiation and multi-cell disturbances in extreme environments. Traditional XOR-based two-dimensional codes have a data width that may vary between 32 and 64 bits, limiting them to single-bit or restricted multi-bit protection. In contrast, the proposed approach provides far wider fault coverage. The verification procedure of block-wise XOR redundancy and cyclic redundancy check (CRC) efficiently addresses issues like burst faults by leveraging the diagonals, parity, and creation of the check-bit. The decoding process employs the encrypted 64-bit input to recover the original 32-bit data, allowing for full data recovery with high accuracy. After we develop and synthesize the design in Xilinx Vivado to test efficiency, we will go into power consumption, LUT utilisation, flip-flop deployment, and I/O efficiency in depth. The experimental findings suggest that the approach provides high resilience at minimal redundancy, together with lightweight protection for memory in aircraft systems, and outstanding dependability. In addition, we have introduced a hybrid CRC-XOR coding and signaling scheme that strikes a balance between error resilience and minimal hardware overhead utilisation, we have ensured that the recovery still returns 32 bits when scaling to 32-bit inputs, and we have explicitly supported burst errors through error correction.

DOI: http://doi.org/10.5281/zenodo.20569606

Alzheimer’s Disease Prediction Using Apache Spark For Large-Scale Data Sets

Authors: Maheswari R, Dr Sajana T, Uma Maheswari G

Abstract: Alzheimer’s disease (AD) is a progressive neurodegenerative disorder that significantly impacts cognitive function and quality of life among aging populations worldwide. Early prediction and diagnosis of Alzheimer’s disease are critical for timely intervention and effective disease management. However, the increasing volume and complexity of healthcare data, including neuroimaging, clinical records, and genetic information, present significant challenges for conventional machine learning approaches. This study proposes a scalable framework that integrates deep learning algorithms with Apache Spark to enable efficient large-scale healthcare data analytics for Alzheimer’s disease prediction. Apache Spark is employed for distributed data preprocessing, feature engineering, and large-scale data management, while deep learning models are utilized to learn complex patterns associated with disease progression. Experimental evaluation demonstrates that the proposed framework achieves high predictive performance while significantly reducing computational overhead compared with traditional approaches. The findings highlight the potential of combining distributed computing and deep learning technologies for scalable and accurate Alzheimer’s disease prediction in modern healthcare environments.

DOI: http://doi.org/10.5281/zenodo.20569845

Ai Driven Predictive Healthcare Analytics Platform

Authors: Neetu Maurya, Ashirwad Kr, Amit, Mohit, Anmol

Abstract: Healthcare systems worldwide are under mounting pressure to shift from reactive treatment models to proactive, data-informed care. Delayed diagnoses, fragmented patient records, and the sheer volume of clinical data produced daily make it increasingly difficult for clinicians to act on time-sensitive information. This paper presents an AI-Driven Predictive Healthcare Analytics Platform that combines machine learning, natural language processing, and real-time patient monitoring to anticipate disease progression and flag high-risk patients before their conditions deteriorate. The platform ingests structured data from electronic health records (EHRs), unstructured clinical notes, laboratory results, and wearable sensor streams, then applies ensemble learning models—including gradient-boosted trees and deep recurrent networks—to generate individualized risk scores. A clinical decision support dashboard surfaces these predictions in plain language, enabling physicians, nurses, and care coordinators to intervene at the right moment without wading through raw data. The system was evaluated across three hospital departments—cardiology, internal medicine, and emergency care—on a retrospective dataset of 84,000 patient records spanning five years. It achieved a predictive accuracy of 91.7% for adverse events within a 48-hour window and reduced clinician alert fatigue by consolidating actionable warnings into a single prioritized feed. The platform is designed to comply with HIPAA and HL7 FHIR standards, runs on commodity cloud infrastructure, and integrates with major EHR vendors through a RESTful API layer. Beyond its immediate clinical utility, the system lays a scalable foundation for population health management, readmission prevention, and personalised treatment planning. This work demonstrates that thoughtfully engineered AI—grounded in real clinical workflows—can meaningfully support human decision-making without replacing the judgment and empathy that define good medicine.

DOI: http://doi.org/10.5281/zenodo.20570254

LeafScan: Plant Leaf Disease Detection Using Convolutional Neural Networks (CNN)

Authors: Hardik Chaturvedi, Aditi Verma, Dr. Raj Kumar

Abstract: Developing countries like India have Agriculture as a major part of their economy. Diseases in crops can damage yields and lead to a reduction in farmer's earnings. Traditional crop disease diagnosis approaches are still laborious and require an expert knowledge. To overcome this issue; this research brings Leaf Scan: an intelligent web application-based Vegetable and Crop Disease detection using Convolutional Neural Network (CNN). The system detects and predicates diseases using images of affected leaves and also suggests the treatment and prevention methods. The system was developed using the Plant Doc dataset and deep learning Mobile Net V2 architecture. The web application brings multi-lingual interaction, disease prediction based on real-time images, scan history management, visual reports, treatment and prevention data, information based on the use case and adaptive browsing interfaces etc. The system has estimated a correctness of 86% on 27 disease classes as per experimental analysis. This project aims to showcase the role of Artificial Intelligence and Deep Learning in augmenting agricultural productivity and crop yield and minimizing crop losses and supporting Precision Farming Technologies.

AI Based Personal Finance Advisor

Authors: Aditi Rajput, Avani Malviya, Dhruv Dekhane, Mahi Ishwe, Prof. Manish Vyas

Abstract: Artificial intelligence is reshaping how individuals access, receive, and act on financial advice. Personal finance— covering budgeting, credit management, investment planning, and retirement preparation—has long been gatekept by cost and expertise. AI-driven systems, from simple calculators to adaptive robo-planners, are changing that. But expanded access does not automatically mean better outcomes. Many platforms reproduce the same incentive misalignments and opacity prob- lems that made traditional human advisors unreliable in the first place, only now at scale. This paper examines the current state of AI in personal finance advice across three angles: (1) a review of machine learning techniques used in credit risk assessment and portfolio optimization, (2) an analysis of trust, personalization, and behavioral risks in deployed platforms, and (3) a proposed five-principle framework—fiduciary duty, adaptive personalization, technical robustness, ethical fairness, and auditability—to evaluate whether these systems genuinely serve users. Experimental results from a supervised ML classifier trained on a Kaggle credit risk dataset show Random Forest achieving 89.25% accuracy and AUC of 0.77. The broader argument is that technological sophistication, while necessary, does not resolve the governance gaps that undermine user trust in AI financial advice.

DOI: http://doi.org/10.5281/zenodo.20573690

Application Of Mathematics In Modern Science And Technology

Authors: T.Narahari

Abstract: Mathematics plays a fundamental role in the development of modern science and technology. Many real-world problems arising in physics, biology, engineering, environmental science, computer science, and medical research can be analysed and understood through mathematical modelling. Mathematical models simplify complex systems and express them in the form of equations, allowing scientists and researchers to study the behaviour of changing quantities with accuracy and clarity. Among the various mathematical tools, differential equations occupy significant place because thy describe relationships involving rates of change and evolving systems. Differential equations are widely used to model natural phenomena such as population growth, radioactive decay, heat transfer, and spread of diseases, motion of falling bodies, electrical circuits, and chemical reactions. In particular, first-order differential equations are highly useful in understanding growth and decay processes where the rate of change of a quantity is proportional to the quantity itslf. These equations help predict future behaviour based on present conditions and are therefore valuable in scientific planning and technological applications. This paper discusses the applications of mathematics in modern science and technology with special emphasis on differential equations and population growth models. The law of natural growth is introduced by assuming that the rate of increase of a population is directly proportional to the existing population. By forming and solving the corresponding first-order differential equation, and exponential growth model is obtained. The paper also presents a real-life application involving bacterial population growth to demonstrate how mathematical techniques are used for prediction and analysis. The study highlights how mathematics supports scientific discoveries, technological innovations, and decision-making processes in modern society. Form predicting epidemics such as Covid-19 to analysing ecological systems and resource planning, mathematical methods provide reliable and efficient solutions. Thus, mathematics serves as the backbone of modern science and technology and continues to contribute significantly to future advancements in research and innovation.

DOI: http://doi.org/10.5281/zenodo.20573767

Health Monitoring Dashboard: Data Analytics & Visualization — Architecture, Implementation, And Future Directions

Authors: Dr Raj Kumar, Ayush Kumar Anand, Harivansh Sharma, Devansh Rohila, Aman Jaiswal

Abstract: The rapid rise in digital health records has sparked a strong need for smart tools that convert unprocessed body data into practical medical decisions. This study introduces the Health Monitoring Dashboard (HMD), a unified web application combining data analytics, interactive visualization, and rule-based health insights. Developed using Microsoft Power BI, the platform leverages its built-in data connectors, DAX (Data Analysis Expressions) for metric calculations, Power Query for data transformation, and AI-powered visuals for predictive analytics. The interface features six components: filtering user conditions, aggregating key performance metrics, displaying time-based trends and cross-data relationships, applying rule-driven health thresholds, and offering an interactive machine learning testing environment via Power BI's integration with Azure Machine Learning. Tests confirm the system accurately handles inputs such as heart rate, blood pressure, sleep habits, steps taken, and physical activity with near-real-time dashboard refresh. Future improvements include direct integration with wearable device APIs (Fitbit, Garmin), secure cloud deployment via Power BI Service and Azure, row-level security for multi-user access, and advanced ML model integration through Azure ML pipelines. The results demonstrate that accessible BI tools can build reliable healthcare analytics systems and offer a replicable blueprint for others.

An Analysis Of Various Control Schemes For Multiarea LFC

Authors: Reena Chauhan, Dr. Charu

Abstract: Automatic generation control, often known as AGC, is necessary in order to ensure that a power system network operates in a consistent manner. The LFC is primarily responsible for controlling the output of the generator in response to changes in the tie-line power and the frequency of the network. Some examples of this include reestablishing the frequency that was intended and exchanging power with other sites within the parameters that were previously established. In this work, the performance of several load frequency control techniques is compared using secondary controllers. These controllers include proportional plus integral plus derivative (PID), integral double derivative (IDD), model predictive control (MPC), and fuzzy cascaded PID. Through the use of a generation rate constraint (GRC), the study is carried out for thermal reheat systems that consist of three areas as well as five areas. A well-known optimization technique known as Big Bang Big Crunch (BBBC) is used in order to ascertain the optimal gains for PID controllers. In order to evaluate the responsiveness of the controllers, a number of different load perturbations in a number of different places have been taken into account. On the basis of settling time, overshoot, and undershoot, it has been able to make comparisons between the results. In every scenario, with the exception of a few peaky oscillations in responses, it is seen that fuzzy cascaded PID works better than other controllers, notably in terms of settling time. This is the case regardless of the conditions. Furthermore, in contrast to conventional controllers, the fuzzy cascaded PID controller has a greater degree of flexibility and durability.

Smart Underground Drainage and Gas Monitoring Systems

Authors: Prof. Kiran Khedkar, Vrunda Ghuge, Anjali Choure, Anamika Khawale

Abstract: Urban drainage systems play a critical role in maintaining sanitation and environmental health. However, manual monitoring of underground drainage systems is inefficient and risky due to the presence of toxic gases such as methane (CH₄), carbon monoxide (CO), and hydrogen sulfide (H₂S). This paper proposes a smart underground drainage and gas monitoring system using IoT and sensor-based technologies. The system uses gas sensors, ultrasonic sensors, and water level sensors integrated with a microcontroller (NodeMCU/Arduino) to continuously monitor drainage conditions. When hazardous conditions are detected, alerts are sent via GSM along with real-time data updates to cloud platforms. The proposed system improves safety, reduces manual effort, and ensures early detection of blockages and gas leaks.

The Silent Threat: Chronic Exposure To Chemical Fertilizers And Its Impact On Human Health

Authors: Lalit Kumar, Darshan Singh, Harivansh Sharma, Deepak Kumar

Abstract: The widespread use of chemical fertilizers has revolutionized agriculture, significantly increasing crop yields and supporting the global food supply. However, the chronic exposure to these chemicals poses a silent but significant threat to human health. This paper explores the various pathways through which chemical fertilizers affect human health, examines the long-term health implications, and suggests potential measures to mitigate these risks. The review synthesizes existing research on the subject, highlighting the need for more comprehensive studies and the development of sustainable agricultural practices that prioritize human health.

DOI: http://doi.org/10.5281/zenodo.20582382

AI-Powered Design Assistant: A Secure, Full-Stack Framework Integrating Prompt Intelligence, Generative AI, And Lossless Canvas Editing

Authors: Khushi Bhandari, Shriraj Uday Shetty, Dr. Jasbir Kaur, Sandhya Thakkar, Ifrah Kampoo

Abstract: Generative AI and professional canvas editing remain largely isolated, forcing creators to juggle multiple tools and lose the benefits of integrated intelligence. We present a full-stack Design Assistant that tightly couples a prompt-expansion engine (GPT-4), a generative image service (DALL·E 3), vision-based asset tagging, and a Fabric.js object-based canvas. Unlike existing platforms that offer only fragmented AI features, our system provides end-to-end prompt refinement, secure back-end AI orchestration, and lossless JSON canvas persistence. We formalise prompt expansion as a structured optimisation that maximises relevance to the user’s intent, and we implement a STRIDE-based threat model to ensure confidentiality and integrity of user assets. In a controlled user study (20 participants), the system improved image relevance from 72% to 94% (Likert scale), reduced task completion time by 38%, and achieved a System Usability Scale (SUS) score of 78.4 (“Good”). We critically compare our approach with commercial AI-design tools (Canva Magic Studio, Adobe Firefly, Figma AI) and openly discuss limitations regarding sample size, API dependency, and scalability. The full architecture, security measures, and evaluation metrics are described in detail, providing a reproducible baseline for future human-AI collaborative design research.

DOI: http://doi.org/10.5281/zenodo.20588710

A Framework For Identifying And Preventing Man-In-The Middle-Attack

Authors: Dr. B. Revathi alias Ponmozhi, Mrs. M. Dhayalini, Kaipu Bhaskar Reddy, Niranjana V

Abstract: Man-in-the-middle (MITM) attacks remain a significant threat to secure communication systems, enabling adversaries to intercept and manipulate data exchanges. This paper presents a framework for identifying and preventing MITM attacks through the integration of cryptographic operations and anomaly detection techniques. The proposed system is implemented in Visual Studio Code (VS Code) and employs AES (Advanced Encryption Standard) with Galois/Counter Mode (GCM) for authenticated encryption, alongside RSA for secure key exchange. These cryptographic primitives ensure confidentiality, integrity, and authenticity of transmitted data. The framework assumes that attackers exploit weak encryption protocols, compromised certificates, traffic analysis to distinguish legitimate communication from malicious activity. Recent advances in cybersecurity, including Al-driven anomaly detection and zero-trust architectures, inform the design of the system. Comparative evaluation demonstrates that the proposed approach achieves higher detection accuracy and lower false-positive rates than conventional intrusion detection solutions, while maintaining minimal computational overhead. The system was implemented using Visual Studio Code and tested through IP based communication between network nodes. It also demonstrates the proposed approach for the securing data transmission and successfully detects potential attacks while maintain the communication integrity. This contribution enhances secure digital communication.

DOI: http://doi.org/10.5281/zenodo.20589123

TrackVerse BMS (Bus Management System)

Authors: C. Sudhakar, K. SaiKrishna, J. Aakhash

Abstract: College bus tracking is a practical problem that students and parents deal with every day you never quite know when the bus will arrive, and that uncertainty adds unnecessary stress. This paper describes a low-cost tracking system we built to solve exactly that. The system uses a Seeed XIAO ESP32S3 microcontroller paired with a u-blox NEO-6M GPS module and a SIM800L GSM module to continuously track a college bus and send location data over the BSNL 2G cellular network. Unlike approaches that rely on Wi-Fi, our system works anywhere there is cellular coverage. Location data including coordinates, speed, altitude, and satellite count is collected every two seconds and sent to a Node.js server hosted on the Railway cloud platform. The server runs a stoppage detection algorithm that flags bus halts when the speed drops below 5 km/h for more than 60 consecutive seconds. A web dashboard built with Leaflet.js shows the live route and highlights each detected stop with timing details. We tested the system on a real campus route over three days and found 100% stoppage detection accuracy with no false positives. The entire hardware cost is under INR 1,500, making this a practical alternative to commercial tracking solutions.

AI-Based Adaptive Indian Sign Language Alphabet Recognition System For Real-Time Gesture Interpretation

Authors: Swastik Manglam, Dr. Poonam Dahiya, Dr. Gaurav Aggarwal

Abstract: Indian Sign Language (ISL) serves as a major communication medium for individuals with hearing and speech impairments; however, its limited accessibility among the general population continues to pose significant challenges. This research presents an AI-based ISL alphabet recognition system designed to accurately identify static hand gestures corresponding to alphabets (A-Z) in real time. This approach integrates CV techniques with DL models to effectively extract hand features and perform robust classification. Unlike conventional systems that rely on strict adherence to predefined gesture patterns, the proposed model emphasizes adaptive recognition by accommodating natural variations in gesture execution across different users and environmental conditions. The system specifically addresses key challenges such as gesture similarity, lighting variations, and differences in hand orientation and appearance. Experimental evaluation demonstrates high recognition accuracy and consistent performance across diverse test scenarios, highlighting the robustness and practical applicability of the approach. Beyond its technical contribution, the system supports educational and assistive applications by enabling intuitive, interactive, and non-memorization-based learning for ISL, particularly benefiting children and beginner users. This work forms a foundational component for the development of a comprehensive ISL translation system and contributes toward advancing inclusive and accessible communication technologies.

DOI: http://doi.org/10.5281/zenodo.20591814

Ai-Based Autism Detection System

Authors: T. Sandhiya, E. Anitha, k. Arunkumar

Abstract: Autism Detect AI is a multimodal deep learning-based computer-aided diagnosis system for Autism Spectrum Disorder (ASD). The system integrates audio and video data to analyze behavioral patterns and classify features associated with ASD. Video data is processed using Convolutional Neural Networks (CNNs), while audio data utilizes Mel-Frequency Cepstral Coefficients (MFCCs) combined with CNNs for classification. The proposed system achieves an accuracy of 95% in identifying ASD in children. This high accuracy demonstrates its potential as an objective and standardized tool for early ASD diagnosis, improving upon subjective clinical evaluation methods. The findings highlight its effectiveness in clinical applications, enabling early intervention and improved developmental outcomes.

DOI: http://doi.org/10.5281/zenodo.20593976

Predictive Resource Estimation For Efficient HPC Cluster Scheduling Using Machine Learning And Slurm Simulation

Authors: Hariprasad Thorve, Pratik Ghadage, Muskan Shaikh, Prafull Barathe, Dr. R.V. Babar, Prof.A.P. Joshi

Abstract: High-performance computing (HPC) environments depend on effective resource allocation to maintain throughput, fairness, and utilization. In practice, many jobs request more memory and execution time than they consume, causing frag-mentation and delays. This paper presents a framework for predictive resource estimation using supervised machine learning, monitoring, and Slurm simulation. Historical job attributes are used to estimate actual memory demand. Linear Regression, Decision Tree, and XGBoost were evaluated. XGBoost achieved the best performance with lowest error and highest score. The study integrates prediction, observability, and simulation to improve scheduling efficiency in institutional HPC clusters.

A Study On the Efforts of GST on Price Stability in Indian Economy with Special Reference to Coimbatore City

Authors: Mr. Lingesh S, Ms. Avanthika R T, Dr. N. Rajendran

Abstract: The implementation of the Goods and Services Tax (GST) has been a landmark reform in the Indian economy, aimed at creating a unified tax structure and ensuring price stability. This study examines how GST influences price stability by eliminating the cascading effect of taxes and streamlining the supply chain. By analyzing various sectors in the industrial hub of Coimbatore, the research highlights how tax rationalization shapes market prices and consumer purchasing power. Findings indicate that while GST has improved long-term price stability in manufacturing and textiles, certain sectors like automobiles in Coimbatore have experienced initial price volatility. The study also discusses the challenges faced by small-scale industries in the region regarding compliance and input tax credits.

DOI: http://doi.org/10.5281/zenodo.20614123

A Study On Introduction to Reselling Business

Authors: Mr. Velmurugan.M, Mr. Jovin.U, Mr.P.Saravana Kumar

Abstract: In the present digital era, business activities have changed significantly due to the rapid growth of technology and internet usage. Traditional business models that required heavy investment, physical stores, and large inventories are gradually being replaced by flexible and low-cost business models. One of the most popular modern business models is the reselling business. Reselling refers to the process of purchasing goods from suppliers or wholesalers and selling them to customers at a profit. The reseller acts as a middleman between the supplier and the customer. With the growth of social media and e-commerce platforms, reselling has become easier and more accessible for people from different backgrounds. Applications such as Instagram, WhatsApp, and Facebook help resellers promote products and communicate with customers. Platforms like Meesho and GlowRoad further simplify the process by offering product catalogs and delivery support. The reselling business is highly popular because it requires low investment and offers flexible working opportunities. Individuals can work from home, choose their own products, and manage their business according to their convenience. It also encourages entrepreneurship by helping people develop skills such as marketing, communication, and customer service. However, the business also faces challenges such as high competition and dependency on suppliers for product quality and delivery. Overall, reselling business has emerged as an important source of income and self-employment in the modern economy.

DOI: http://doi.org/10.5281/zenodo.20614302

A Study on Role of Social Media Marketing in Brand Building

Authors: Mr.Ahamed Nafih.N.R, Mr. Mubarak. S, Dr. N. Rajendran

Abstract: As digital landscapes evolve; the methodology of brand building has undergone a paradigm shift from traditional mass media to interactive social platforms. This study investigates the factors that encourage brands to leverage social media for identity construction and the elements that drive consumer engagement. The research identifies key pillars—authenticity, community engagement, and visual storytelling—that facilitate brand resonance. Primary data was collected from a sample of 65 respondents to analyse consumer perception of branded social content. Findings indicate that consistency and two-way communication are the primary drivers of brand trust. Furthermore, the study highlights how social media serves as a "one-stop solution" for brand-consumer relationship management in the modern era .building on these insights ,the study emphasizes that social media marketing not only enhances brand visibility but also strengthens long term brand equity .Platforms such as Instagram ,Facebook ,x, and YouTube enable brands to humanize their identity by sharing real time content, behind the scenes narratives, and user generated experiences .This participatory environment empowers consumers to co-create brand meaning, thereby increasing emotional attachment and loyalty. Additionally, data driven targeting and analytics allow marketers to personalise messages, improving relevance and recall among diverse audience segments’ research further observes that responsiveness to feedbacks, reviews, and grievances significantly influences consumer satisfaction and brand credibility. Hence, social media marketing has become an indispensable tool for sustainable brand building in the contemporary digital ecosystem.

DOI: http://doi.org/10.5281/zenodo.20614392

A Study on Financial Analysis of Private and Public Sector Banks in India

Authors: Ms. S. Priyadharshini, Mr. P. Hariharan, Dr. N. Rajendran

Abstract: The banking sector plays a crucial role in the economic development of India by mobilizing savings, providing credit, and supporting financial stability. This study focuses on the financial analysis of selected private and public sector banks in India to evaluate their performance in terms of profitability, liquidity, solvency, and operational efficiency. By comparing two private sector banks and one public sector bank using financial ratios and secondary data sources, the study aims to identify performance differences and understand the strengths and weaknesses of each banking segment.

DOI: http://doi.org/10.5281/zenodo.20614488

Twitter Data Analysis For Suicidal Intent

Authors: Vivek Nagargoje, Tushar Tayade, Prathmesh Dhamale, Kshitij Ghumare

Abstract: Early identification of suicidal intent on social media is vital for effective suicide prevention. This paper proposes a robust system for detecting suicidal intent from Twitter data using a Support Vector Machine (SVM) classifier as the core algorithmic component. The system pipeline includes text preprocessing, TF-IDF-based feature extraction, sentiment score enrichment, and SVM-based binary classification. The SVM model with RBF kernel is trained and evaluated on a labeled Twitter dataset, achieving an accuracy of 94.2%, precision of 92.5%, recall of 91.8%, and F1-Score of 93.3%. The proposed SVM-based approach offers a practical balance between classification accuracy and computational efficiency, making it well-suited for real-time deployment in mental health monitoring systems.

DOI: http://doi.org/10.5281/zenodo.20621545

A Study on Consumers Buying Behaviour Towards Fmcg Products with Reference to Coimbatore District

Authors: Mr. Mohammed Suhail, Mr. Anantheeswaran

Abstract: This study highlighted that consumers place greater value on the quality of fast-moving consumer items when making purchases from particular brands. This study determines how much different factors affect the respondents' decisions to buy FMCG items. FMCG branding has become an essential element of consumers' daily lives. Every day, consumers are actually faced with hundreds of companies. This was accomplished by determining the primary factors of branding, quality, and the four Ps (price, packing, promotion, and purity). The study found that customers rely on branding and product quality, with the remaining factors having the least influence. Even while rural customers use their popular branded items across all product categories and spend a significant amount of their income on them, these products are now typically consumed by all cultures. While reducing the possibility that customers may favour certain companies because they are well-known to them or because of ads. This study also shows that, despite their modest participation with some items, customers form their attitudes and behaviours toward FMCG brands. Despite a number of complaints, it was effective in instilling brand ideals in the brains of its customers.

DOI: http://doi.org/10.5281/zenodo.20621920

A Study on Customer Awareness, Satisfaction and Investment Behaviour Towards Post Office Savings Schemes with Special Reference to Coimbatore City

Authors: Mr.Mohammed Ejaz.M, Mr. Kalaiyarasan.A, Dr. Rajendran.N

Abstract: Post Office Savings Schemes are one of the most trusted investment options in India. These schemes are supported by the Government and provide secure and stable returns. In a developing economy like India, savings play a vital role in ensuring financial stability. People from different income groups prefer these schemes due to their low risk and accessibility. In Coimbatore City, a mix of urban and semi-urban population makes it an ideal place to study customer behaviour. Post Office Savings Schemes are one of the most trusted investment options in India. These schemes are supported by the Government and provide secure and stable returns. In a developing economy like India, savings play a vital role in ensuring financial stability. People from different income groups prefer these schemes due to their low risk and accessibility. In Coimbatore City, a mix of urban and semi-urban population makes it an ideal place to study customer behaviour. Post Office Savings Schemes are one of the most trusted investment options in India. These schemes are supported by the Government and provide secure and stable returns. In a developing economy like India, savings play a vital role in ensuring financial stability. People from different income groups prefer these schemes due to their low risk and accessibility. In Coimbatore City, a mix of urban and semi-urban population makes.

DOI: http://doi.org/10.5281/zenodo.20622404

A Study on Consumer Perception Towards Rapido in Coimbatore City

Authors: Ms.Madhumitha Mondal.P, Mr. Manikanda Eswaran.A, Ms.Sri Ranjani.A. R, Mrs.Jeya Padma Deepa.I

Abstract: This study investigates customer preference, awareness, and satisfaction towards Rapido bikes, a bike-sharing service aimed at providing convenient and sustainable transportation solutions in urban areas. Through a combination of quantitative surveys and qualitative interviews, data was collected from a sample of Rapido bike users in various cities. The study examines factors influencing customer preference for Rapido bikes, including convenience, cost-effectiveness, and environmental sustainability. Additionally, it explores levels of awareness regarding Rapido bike services and their perceived benefits and challenges. Furthermore, the study assesses customer satisfaction with Rapido bikes, focusing on aspects such as service reliability, safety, and overall user experience. Findings from the study provide valuable insights into the factors driving customer preference and satisfaction towards Rapido bikes, as well as opportunities for enhancing awareness and improving service quality. These insights can inform strategic decision-making and policy interventions aimed at promoting the adoption of bike-sharing services as a viable mode of urban transportation, contributing to sustainable mobility and reduced congestion in urban areas.

DOI: http://doi.org/10.5281/zenodo.20622506

A Study on Digital Banking Services in Indian Bank with Special Reference to Coimbatore City

Authors: Mr.Abdul Mahamooth.A, Ms. Afsana.A.H, Mrs.Chithra. B

Abstract: Digital banking has become an important part of the modern banking system due to rapid technological development. This study focuses on digital banking services provided by Indian Bank in Coimbatore city. The main objective is to analyse customer awareness, usage, and satisfaction towards digital banking services. The study is based on primary data collected from 100 respondents using a structured questionnaire. The findings reveal that most customers prefer digital banking due to convenience and timesaving, but some face issues such as security concerns and technical problems. The study concludes that improving awareness and security measures can enhance customer satisfaction and digital banking usage.

DOI: http://doi.org/10.5281/zenodo.20622653

A Study on Effectiveness of Titan in Tamilnadu

Authors: Ms. Subhashini.A, Ms. Archana.R, Dr. Prabhakaran.K, Mr.Saravana Kumar.P

Abstract: The Indian branded consumer goods market has witnessed significant growth over the past few decades, with the lifestyle and accessories segment emerging as a highly competitive and dynamic industry. Titan Company Limited has played a pioneering role in transforming the organized watch, jewelry, and eyewear markets in India. This article examines the effectiveness of Titan in Tamil Nadu, one of its most important and mature markets. The study analyzes Titan’s brand strategy, product portfolio, pricing, distribution network, promotional practices, and customer satisfaction levels within the state. Emphasis is placed on understanding how cultural alignment, trust, innovation, and service quality contribute to Titan’s sustained market leadership. Using secondary data, case insights, and existing literature, the study highlights the factors responsible for Titan’s strong brand equity and evaluates its effectiveness in meeting consumer expectations in Tamil Nadu. The findings reveal that Titan’s localized strategies, ethical branding, and customer-centric approach have significantly enhanced its effectiveness and long-term competitiveness in the region.

DOI: http://doi.org/10.5281/zenodo.20622832

Expense Tracker Web Application: Design And Development Using Angular, ASP.NET Core, And MySQL

Authors: Dr. Rajkumar, Vipin Kushwaha, Vasu Aggarwal, Vishal Singh, Ujjawal Charnotia

Abstract: Effective personal financial management is a critical life skill that many individuals struggle to practice consistently. Conventional approaches such as paper ledgers and basic spreadsheet files are error-prone, difficult to analyze, and inadequate for modern financial needs. This paper presents the design, implementation, and evaluation of a web-based Expense Tracker Application intended to help users record, categories, and visualize their daily financial transactions in a structured and user-friendly manner. The proposed system is built on a three-tier architecture comprising Angular 14 on the client side, ASP.NET Core on the server side, and a MySQL relational database for persistent data storage. Communication between the front-end and back-end layers is facilitated through RESTful Application Programming Interfaces (APIs) transmitted over the Hypertext Transfer Protocol (HTTP). The application supports secure user authentication using JSON Web Tokens, category-based expense classification covering food, transportation, entertainment, utilities, and shopping, and dynamic report generation through interactive charts and tabular summaries. The system was designed with responsiveness in mind to perform consistently across desktop computers, laptops, tablets, and mobile devices. The study adopts the Software Development Life Cycle (SDLC) model encompassing requirements analysis, system design, implementation, and systematic testing. Evaluation results demonstrate that the proposed application significantly simplifies personal budgeting, enhances financial awareness, and addresses key limitations of commercially available tools. The paper further discusses the system architecture, database design, API structure, security mechanisms, and future enhancement directions.

AI-Powered Smart Digital Library With Conversational Interface For Personalized Knowledge Discovery

Authors: Atul Parjapati, Manav Kumar

Abstract: Rapid growth in the availability of digital aca-demic libraries is increasing information retrieval complexities. Keyword-based mechanisms cannot understand user queries and do not support any form of personalization, thus making knowledge acquisition cumbersome. This research paper presents an intelligent AI-enabled digital library model that combines conversational intelligence with a rule-based ranking system. Different from traditional machine learning models, the pre-sented approach does not involve complex and computationally expensive learning but remains accurate and efficient at the same time. According to experimental tests, the new method produces better precision (0.84), recall (0.81), and faster response time (1.15 seconds).

DOI: http://doi.org/10.5281/zenodo.20624585

AI-Powered Misinformation Detection System Attrition

Authors: Bharathi Panduri, P K Abhilash, Dr. Y J Nagendra Kumar, Kaliveni Naveen, Alluru Manoj, Patha Shiva Anurag

Abstract: The widespread dissemination of misinformation through social media, news platforms, and messaging applications has become a major challenge in the digital age, particularly in India. False and misleading information can trigger public panic, financial fraud, health risks, and social instability. Existing fact-checking solutions are often manual, time-consuming, or inaccessible to common users, creating a need for an intelligent and real-time misinformation detection system. This project, Verify & Learn, presents an end-to-end solution that leverages Artificial Intelligence (AI), Machine Learning (ML), and Generative AI (GenAI) to detect, explain, and educate users about misinformation. The system analyzes digital content such as news headlines, articles, social media posts, URLs, and messaging text using machine learning and natural language processing models. These models classify content into categories such as true, misleading, or false across multiple domains including health, politics, finance, science, social media, and online scams. To improve reliability and transparency, the system integrates trusted external sources through fact-checking and news APIs, enabling evidence-based verification with proper citations. Generative AI plays a crucial role in enhancing explainability and user understanding by converting analytical results into clear, human-readable explanations. It also generates educational insights, domain-specific awareness tips, and short learning content that help users recognize common misinformation patterns and manipulation techniques. The proposed system is implemented as a web application and a browser extension, allowing real-time verification during browsing and deeper analysis through detailed reports and shareable PDFs. By combining predictive ML models with explainable and educational Generative AI, this project promotes digital literacy, informed decision-making, and responsible information sharing. The solution demonstrates a scalable and user-centric approach to combating misinformation in real-world digital environments.

DOI: http://doi.org/10.5281/zenodo.20625107

Disease Prediction from Symptoms using Machine Learning

Authors: Assistant Professor Dr Rajkumar, Mehreen, Abiha kazmi, Km Ilma, Apeksha Kaushik, Kritika

Abstract: The goal of this project is to develop a smart system that can forecast illnesses like typhoid, dengue, and malaria by analyzing patient-reported symptoms. It uses three different machine learning algorithms—Random Forest, Decision Tree, and Support Vector Machine—to identify diseases from a predefined list of symptoms and the corresponding diagnosis. Every algorithm has undergone rigorous training and testing to guarantee its precision and dependability in forecasting. When it comes to disease prediction, the Random Forest method outperforms the Decision Tree and Support Vector Machine models. By accurately identifying the relevant illness based on the user's symptoms, it consistently yields reliable results, making it the most effective model in our system.

DOI: https://doi.org/10.5281/zenodo.20627516

E-Waste Management Portal

Authors: Assistant Professor Dr. Raj Kumar, Nikhil, Aditya, Arushi, Chandni

Abstract: The rapid growth of technology and the rise in electronic device use have led to an increase in electronic waste (e-waste) worldwide. Improper disposal of e-waste can cause serious environmental and health risks due to hazardous materials like lead, mercury, and cadmium. Many people do not know how to dispose of e-waste correctly and often struggle to find authorized collection centers. To tackle this problem, this research suggests an E-Waste Management Portal, a web-based platform that aims to simplify electronic waste collection and management. The proposed system allows users to create accounts, log in securely, find nearby e-waste collection centers using Google Maps API, and schedule pickup requests for e-waste from their location. Users can also track the status of their pickup requests in real-time. On the administrative side, the portal offers a main dashboard for managing user requests, monitoring pickup activities, and efficiently processing collection requests The system uses HTML for the front-end interface, PHP for server-side processing, and Google Maps API for location services. By providing an easy-to-use solution, the portal promotes responsible e-waste disposal, boosts recycling efforts, and supports environmental sustainability. This proposed solution shows how web technologies can help with effective e-waste management and encourage eco-friendly practices in communities.

DOI: http://doi.org/

Antimicrobial Activity of Turmeric against Bio-Film Forming Staphylococcus Aureus Isolated from Dairy Milk

Authors: Shudanshu Sharma, Ravindra Kumar Jain

Abstract: Staphylococcus aureus is a well-known pathogen capable of producing enterotoxins during bacterial growth in contaminated milk, causing mastitis in dairy cattle and the ingestion of such preformed toxins is one of the major causes of food poisoning around the world. The capacity of S. aureus to produce biofilms increases its antibiotic resistance, posing a considerable challenge to dairy hygiene and public health. This research explored the antimicrobial activity of turmeric (Curcuma longa) extracts towards biofilm-producing S. aureus isolated from dairy milk. Milk samples were collected from dairy farms and screened to microbiological analysis for the isolation and identification of S. aureus by using selective culture media, morphological, and biochemical methods. The biofilm-forming ability of the isolates was determined by using Congo Red Agar method and the microtiter plate assay. Ethanolic turmeric extract was prepared using appropriate maceration extraction methods and tested for antimicrobial activity

DOI: https://doi.org/10.5281/zenodo.20629957

A Review of Virtual ATM Architectures and Emerging Technologies for Secure Digital Financial Services

Authors: Research Scholar Sahazad Ahmad, Assistant Professor Yashveer Singh

Abstract: The rapid digital transformation of the banking sector has significantly influenced the development of intelligent and secure financial transaction systems. Traditional Automated Teller Machines (ATMs), although effective for self-service banking, face multiple challenges related to security vulnerabilities, physical infrastructure dependency, operational cost, and limited accessibility. This review paper presents a comprehensive analysis of Virtual ATM systems and their future prospects in modern digital banking ecosystems. The study explores the evolution of ATM technologies from conventional card-based systems to advanced cardless, contactless, cloud-connected, and AI-enabled Virtual ATM frameworks. Various technologies such as Near-Field Communication (NFC), QR-code authentication, biometric verification, One-Time Passwords (OTP), blockchain integration, cloud computing, and Artificial Intelligence (AI)-based fraud detection are critically examined. The paper also investigates ATM virtualization architecATMaaS), and cloud-managed banking infrastructures. Furthermore, the review highlights the role of biometric authentication techniques such as fingerprint, facial recognition, retina scanning, and multi-modal biometric fusion in enhancing transaction security and user convenience. Applications of Virtual ATM systems in smart banking, financial inclusion, accessibility support, and cashless transactions are also discussed. Additionally, major challenges including cybersecurity threats, privacy concerns, infrastructure limitations, biometric data protection, and regulatory compliance are analyzed. Comparative analysis demonstrates that AI-enabled and blockchain-based Virtual ATM systems offer improved scalability, operational efficiency, transparency, and security compared to traditional ATM infrastructures. The study concludes that Virtual ATM technology represents a promising direction for future intelligent banking systems and secure digital financial services.

DOI: http://doi.org/

Cloud Native Decision Support Framework for Crop Stress Monitoring Using Random Algorithm

Authors: Dr. Pritesh Patil, Vinay Basargekar, Shraddha Thorbole, Saurabh Rai

Abstract: Satellite imagery has become an essential component of modern precision agriculture, enabling large-scale observation of crop health conditions. However, many existing monitoring solutions primarily provide vegetation indices, stress maps, or threshold-based alerts that can be difficult for farmers to interpret and act upon. To address this limitation, this study presents a cloud-native decision-support framework that integrates multispectral satellite observations, weather information, machine learning, and explainable recommendations. The proposed framework uses key vegetation indicators, including NDVI, NDWI, and SAVI, from Sentinel-2 satellite imagery and combines them with environmental parameters such as rainfall, temperature, and humidity. A Random Forest classifier is utilized to categorize crop conditions into four stress levels: healthy, mild, moderate, and severe. To enhance transparency and usability, an explainability module identifies the primary factors influencing each prediction and translates them into actionable recommendations, such as checking irrigation systems, assessing nutrient deficiencies, or maintaining routine field monitoring. The system follows a modular cloud-native architecture including data retrieval, preprocessing, feature engineering, model prediction, and dashboard visualization. The key contribution of the proposed framework is its ability to convert analytical results into actionable insights. Beyond detecting crop stress levels, it provides explanations for the underlying causes and recommends suitable management practices, enabling farmers to make informed decisions.

DOI: http://doi.org/

Comparison Study on Different Neural Network Techniques for Kidney Stone Diagnosis

Authors: Parul Tyagi, Dr. Brij Mohan Singh

Abstract: Kidney stone disease — clinically referred to as nephrolithiasis — remains one of the most painful and widely encountered urological conditions worldwide. Catching it early and getting the diagnosis right can dramatically change a patient's care pathway and reduce the financial strain on healthcare systems. In this study, we take a close, side-by-side look at five machine-learning techniques that have shown promise for automated kidney-stone detection: the Multilayer Perceptron trained with Back Propagation (MLP-BPA), Radial Basis Function (RBF) networks, Learning Vector Quantization (LVQ), Support Vector Machines (SVM), and Deep Convolutional Neural Networks (CNN). All experiments are run on a standardised clinical dataset using WEKA 3.7.5 and Python, with each model assessed on accuracy, sensitivity, specificity, and F1-score. The features fed into every model include creatinine and BUN levels, CT-scan findings, kidney size and contour, and several urinary markers. Among the classical approaches, SVM came out on top with 93.6% accuracy, while MLP-BPA was close behind at — a CNN — pushed accuracy to 96.1% when adequate training images were available. Beyond the raw numbers, we discuss what each architecture actually trades off in practice: how hard it is to train, how transparently it reaches its decisions, and whether a busy nephrology clinic could realistically deploy it. Our hope is that this comparison gives clinicians and AI researchers a clear, honest basis for choosing the right tool for kidney stone diagnosis.

DOI: https://doi.org/10.5281/zenodo.20637927

Student Budget Planner: A Web-Based Financial Management System Using Machine Learning

Authors: Dr Raj Kumar, Manvi Talwar, Chhavi, Misbah Rani, Aastha Kashyap

Abstract: Financial management is an essential aspect of student life, yet many students face difficulties in tracking expenses, maintaining budgets, and planning future spending. This research presents the development of a Student Budget Planner, a web-based financial management application designed to assist students in managing their income and expenses effectively. The system is developed using Python and integrates Streamlit for the user interface, SQLite for database management, Pandas and NumPy for data processing, Matplotlib for visualization, and Scikit-learn for machine learning-based expense prediction. The application enables users to record transactions, categorize expenses, analyze spending patterns through graphical representations, and forecast future expenses using a Linear Regression model. The proposed system offers a simple, interactive, and intelligent solution for personal financial management while providing practical exposure to software development, data analytics, and machine learning concepts. The results demonstrate that the application successfully improves financial awareness and budgeting efficiency among students.

Design And Performance Of Bacteria-Based Self-Healing Concrete

Authors: Sandeep Singh, Sougata Chattopadhyay

Abstract: Concrete cracking is one of the major causes of structural deterioration and durability loss in infrastructure systems. Conventional repair techniques are expensive, labor-intensive, and often ineffective in inaccessible locations. This study investigates the development and performance of bacteria-based self-healing concrete utilizing microbiologically induced calcite precipitation (MICP). Bacillus subtilis spores were encapsulated within lightweight aggregates and incorporated into concrete mixtures to promote autonomous crack healing. The experimental program evaluated fresh concrete properties, compressive strength, water absorption, crack-healing efficiency, and durability performance. Results indicated that the bacterial concrete exhibited comparable mechanical properties to conventional concrete while demonstrating significant self-healing capability. Crack widths up to 0.45 mm were effectively sealed within 28 days of water exposure. Water permeability was reduced by approximately 82%, while compressive strength increased by 12% compared with the control mixture. The findings confirm that bacteria-based self-healing concrete represents a sustainable and cost-effective solution for extending infrastructure service life and reducing maintenance requirements.

DOI: http://doi.org/10.5281/zenodo.20637819

Analysis of Triboelectric Tile for Energy Harvesting from Walking

Authors: Satish Kumar, Geetanjali Kale, Rushikesh Bankar, Rajeev Kumar

Abstract: Harvesting mechanical energy from human walking by triboelectric tile is an effective approach for a sustainable, maintenance free, and green power source. The triboelectric tile is capable of harvesting vibrational energy. In this paper, a simplified model of triboelectric tile system using simple coupling of commercial polyvinyl fluoride (PVDF) with a thin copper film is presented. Based on the simplified model, Real-time output characteristics of triboelectric tile at different values of resistance are derived. The output characteristics of triboelectric tile are validated with the existing literature.

DOI: http://doi.org/10.5281/zenodo.20637966

CropGuard: An AI-Based Crop Disease Detection And Farmer Advisory System For Smallholder Farming In Liberia And Sub-Saharan Africa

Authors: Amara Fuad Nyei, Dr. Vaibhav Bhushan Tyagi

Abstract: This paper presents CropGuard, a full-stack artificial intelligence-powered diagnostic and advisory system designed to bridge the critical crop disease knowledge gap confronting smallholder farmers in Liberia and Sub-Saharan Africa. Agriculture constitutes approximately 36 percent of Liberia’s gross domestic product, yet annual pathogen-induced yield losses of 30 to 80 percent and a national extension officer-to-farmer ratio of 1:35,000 leave the overwhelming majority of the country’s 338,492 farming households without timely diagnostic support. CropGuard integrates a MobileNetV3-Small deep learning backbone-trained on PlantVillage and Makerere University AI Lab datasets – to classify 16 disease categories across five core Liberian staple crops: Bean, Cassava, Corn, Potato, and Tomato. Google Gemini 3.0 Flash Preview synthesises diagnostic outputs into localised agronomic remediation advice across six languages, with Text-to-Speech delivery serving the 64.5 percent of rural female-headed households with no formal schooling. Technical evaluation confirms a validation accuracy of approximately 86 percent with inference latency below 200 milliseconds, demonstrating the system’s viability as a low-cost, scalable solution aligned with Liberia’s National Agricultural Development Plan (NADP 2024–2030) and the continent-wide imperative for digitally-enabled food sovereignty.

DOI: http://doi.org/10.5281/zenodo.20639226

Geometric Terracotta Façades for Passive Cooling in Hot- Dry Climate

Authors: Ar. Suman Sharma, Shavni jain

Abstract: This study compares different terracotta façade geometries to find which shape gives the best passive evaporative cooling in hot–dry climates. Hot–dry regions usually have very high temperatures (above 40°C) and very low humidity (below 25%), which increases the need for cooling in buildings. Using mechanical air-conditioning consumes large amounts of energy, so passive cooling methods are important for sustainable architecture. Evaporative cooling works when water evaporates from a surface and absorbs heat from the surrounding air, reducing temperature. Terracotta is a suitable material for this purpose because it is porous (20–30% porosity), meaning it can absorb water through capillary action and slowly release moisture for continuous evaporation. It also has good thermal mass, which helps in reducing temperature fluctuations between day and night. The research studies different geometric shapes of terracotta façade modules such as square, hexagonal, triangular, cylindrical, grooved, perforated, and pleated forms. Geometry plays an important role in cooling performance because it affects surface area, airflow movement, shading, and water retention. Shapes with higher surface area and deeper recesses allow more water to stay on the surface and increase evaporation, resulting in better cooling. The study compares the performance of these shapes using parameters such as Surface Area to Volume ratio (SA/V), ventilation efficiency (ACH), shading coefficient (SC), and water retention time. Findings show that simple planar shapes like square tiles provide basic cooling, while complex shapes perform better. Cylindrical forms improve airflow and provide consistent cooling. Pleated or capillary geometries give the best results because they create larger surface area, better shading, and improved water distribution. Results from case studies show that pleated terracotta façades can reduce surface temperature by about 10–13°C, while cylindrical systems provide 9–11°C cooling. Perforated and hexagonal geometries show moderate performance. Based on comparative analysis, pleated or capillary geometries achieve the highest Evaporative Cooling Index (ECI = 10.0) and are considered the most suitable option for passive cooling in hot–dry climates. The study concludes that terracotta façades with optimized geometry can significantly reduce heat gain and cooling energy demand in buildings. Pleated or capillary designs are recommended for new façade systems, while cylindrical and grooved forms are suitable alternatives depending on project requirements. This research helps architects select façade geometries based on performance rather than only aesthetics, contributing to climate-responsive and sustainable building design.

Hybrid Work Models And Their Effect On Organizational Culture, Productivity, And Talent Sustainability

Authors: Neethu P Chandran, Vishwas G

Abstract: The onset of the pandemic has led to a radical experimentation on remote working, thus transforming fundamental beliefs of the location, time, and means for conducting business. Now that companies start to shift from the necessity-based remote working into hybrid working, management faces important questions regarding the future impact of such a change on organizational culture, productivity, and employee retention. This paper will provide empirical evidence on various hybrid working models in terms of their adoption among 50 organizations from the United States, Europe, and Asia-Pacific region (totalling over 50,000 employees). By using mixed methods including surveys among n=5,000 employees, in-depth interviews (n=150 managers), and productivity metrics (collaboration tool usage, completion rates for projects), the paper will examine the productivity and effectiveness of four hybrid working models: (1) Remote First, (2) Structured Hybrid (in-office days fixed), (3) Flexible Hybrid (flexibility left to employees), and (4) Office-Centric. Results indicate that Structured Hybrid provides higher productivity (+8.2% compared to pre-pandemic) and lower turnover (by 18%), while Remote First ensures better access to talents, yet poor mentoring. Flexible Hybrid is best received by employees but entails higher coordination costs.

DOI: http://doi.org/10.5281/zenodo.20666545

PalmPilot: A Touchless Gesture Control System

Authors: Nitin Dhawas, Ritesh Sirpor, Yash Patil, Jamal Siddiqui

Abstract: Contemporary advances in computer vision and edge-optimized neural inference have established a practical pathway for deploying high-fidelity hand-tracking pipelines on commodity hardware, enabling entirely touchless human-computer interaction. This paper presents a vision-based, touchless gesture-controlled operating system interface that converts natural hand movements captured by a standard RGB webcam into a comprehensive set of OS-level input events. The proposed framework integrates Google's MediaPipe Hands for real-time 21-landmark detection, OpenCV for acquisition and preprocessing, a discrete-time Kalman filter for tremor suppression, and PyAutoGUI/pynput for platform-native event dispatch. Nine interaction primitives are supported: cursor movement, left/right click, double-click, drag-and-drop, bidirectional scrolling, three-finger Task View invocation, and pinch-to-zoom. A frame-persistence

DOI: http://doi.org/

Privacy Leakage Through Background App Permissions in Android Devices

Authors: Sushovan Chandra, Omar Faruk Molla, Pritam Samanta, Suchandra Bharati, Barsha Maity, Susmita Bhaskar

Abstract: The rapid expansion of Android smartphones has transformed the way individuals communicate, work, and access digital services. However, the widespread adoption of Android applications has introduced significant privacy concerns, particularly regarding background permissions that allow applications to access sensitive user data without active user interaction. Many applications request permissions such as location access, microphone usage, camera access, storage management, and contact retrieval. While these permissions are often justified by application functionality, they can also be exploited to collect personal information continuously in the background. This study investigates privacy leakage resulting from background app permissions in Android devices. The paper analyzes Android's permission architecture, explores common privacy threats, examines real-world examples of data leakage, and evaluates current mitigation techniques. Findings indicate that excessive permission requests, insufficient user aware

DOI: https://doi.org/10.5281/zenodo.20671433

Neuromarketing Analytics For Predicting Consumer Purchase Intent In Digital Marketplaces

Authors: Dr. Raja Sekhar Koduru, Saranya P K

Abstract: As a result, the competition for consumers' attention has been increased because of the rapid development of digital marketplaces. At the same time, the current approaches to consumer analytics are based on the use of self-reported measures and do not reflect subconscious processes. Neuromarketing or neuroscience marketing can be described as the application of neuroscientific methods to understanding consumer behavior. In other words, neuromarketing can be used to investigate the mechanisms of making purchasing decisions. This paper introduces the neuromarketing analytics framework based on EEG, ET, and GSR technologies to predict purchase intent in digital marketplaces. Based on data collected from 120 participants who were shown e-commerce product listings, spectral EEG features (theta, alpha, beta, and gamma bands), ET measures (fixation duration, saccade amplitude, and pupil dilation), and GSR phasic responses have been extracted. The proposed deep learning model combines TCN and multi-head attention architecture and achieves 89.2% accuracy in predicting purchase intent. The performance of the proposed model significantly outperforms unimodal baseline models (EEG-based: 76.4%; ET-based: 78.1%; GSR-based: 71.2%). The most significant predictors of purchase intent are found to be gamma band power (30-45 Hz) during product exposure and pupil dilation change.

DOI: http://doi.org/10.5281/zenodo.20672219

Ai Driven Quality Care Assessment System

Authors: Dr. Rajkumar, Vaidik Tyagi, Somaydeep, Anish Afridi , Shubham Kumar, Md Atakarim

Abstract: The Ai Driven Quality care Assessment System is a revolutionary web application that leverages machine learning algorithms to predict patient health risks with 97% accuracy, categorizing them as Low, Medium, or High. This system utilizes a trained Random Forest classifier and is built using Python, Flask, and scikit-learn, with a modern responsive design using Bootstrap 5 and interactive visualizations powered by Plotly.js. The system provides real-time risk assessment, interactive data visualizations, and clinical recommendations, making it an ideal application for hospital emergency departments, clinic patient monitoring systems, and telemedicine platforms. The system's advanced analytics dashboard provides comprehensive model performance metrics and feature importance analysis, while its RESTful API enables seamless integration with other healthcare systems. As demonstrated in Fig 1, the system's architecture showcases professional software development practices, and its applications extend to educational settings as a perfect final year project for computer science students, machine learning course demonstrations, and healthcare informatics research projects, as supported by previous studies [1], [2].

DOI: http://doi.org/10.5281/zenodo.20672662

Automatic Solar Seed Sower

Authors: Aparna Shendkar, Sejal S. Dhanavade, Aditya A. Dhalkari, Mahadev B. Dhanba, Om P. Dhande

Abstract: This project describes the design and implementation of a solar-powered seed sower machine for small and mid-scale farmers. In the machine, the solar power acts as the driving force, and the system is eco-friendly and economically feasible. In the proposed system, the use of solar power will reduce the dependency on human labor and increase the efficiency of the seed sowing process while ensuring equal distribution of seeds. In addition, the solar-powered seed sower helps automate the process of seed sowing, and the seeds will be planted at the correct distance and depth, allowing for efficient crops to be obtained and optimizing the amount of seeds used during the process. Moreover, the use of the solar-powered seed sower facilitates the farmer in terms of reduced effort and cost associated with planting and maintaining crops. Finally, the solar seed sower machine is simple to construct and handle, and the system requires low maintenance.

DOI: http://doi.org/10.5281/zenodo.20672737

Intelligent Traffic Monitoring Using Machine Learning for Violation Detection and License Plate Extraction

Authors: Associate Professor K.V.S.S Ramakrishna, Gurram Anantha Lakshmi, Nalleboina Sravnathi, Naru Venkata Gayathri, Puli Durga

Abstract: The rise of city populations and vehicle numbers per city have led to serious traffic jams and frequent disobeying of traffic laws. Most of the time, authorities use traditional traffic monitoring systems that require manually checking or using sensors to detect the problem, and these may not work well and can be easily forgotten by humans. This research aims to design an intelligent traffic monitoring system using the latest ML and computer vision techniques to identify different traffic violations and get the vehicle license plate number(s) without any human intervention and in real time. The system not only detects red-light running, speeding, and lane crossing but also identifies the responsible vehicles using object detection algorithms (like YOLO or Faster R-CNN) on the video footages captured by the surveillance cameras. The use of OCR for reading license plates after getting the region of the license plate by image processing using deep learning models is also proposed in this system The proposed solution significantly upgrades faithful, timely, and effortless traffic monitoring, thus providing the road safety officers with the necessary tools to keep the roads safe, reduce accidents, and increase the execution of the law with a minimum of human effort.

DOI: https://doi.org/10.5281/zenodo.20674596

Integration of IoT and Remote Sensing for Accurate Crop Yield Estimation

Authors: Associate Professor V.Pavani,, Kongara Triveni,, Evuri Bhavya Sri,, Palapala Pranathi,, Jogi Keerthi

Abstract: One of the biggest challenges fed by climate change and environmental issues are global food production. This has led to the demand of a sustainable food source as the main driver for transforming agriculture into more eco-friendly practices. In this context, the research emphasizes IoT-RS Integrated Smart Yield Prediction Model (IRSYPM) is a single intelligent system to combine the Internet of Things (IoT) sensor networks and Remote Sensing (RS) data for yield prediction and precision agriculture assistance. The model employs current data that IoT devices record as they monitor the soil, air, and nutrient content. At the same time, remote sensing from either drone or satellite is used to derive the multispectral vegetation indices like NDVI, EVI, and LAI. All these different data inputs come together on a cloud platform that uses machine learning algorithms such as Random Forest, Support Vector Regression (SVR), and Artificial Neural Networks (ANN) to predict yield outcomes.The model IRSYPM is structurally ₂ emissions by 19%. Such effects are a great contribution to eco-friendly and data-driven agricultural practices.Essentially, the IRSYPM scheme is a large-scale, secure, and intelligent framework that revolutionizes the conventional agri-food system into a smart, resilient, and environmentally friendly one.

DOI: https://doi.org/10.5281/zenodo.20674780

Fitcore : An AI-Driven Intelligent Gym Training System

Authors: Priyanshu Patidar, Sachin Prajapat, Samay Jain, Satyam Thakur, Dr. Brajesh Chaturvedi

Abstract: This paper presents a systematic review of contemporary health informatics architectures, focusing on the paradigm shift from static fitness applications to dynamic, generative ecosystems. Traditional fitness platforms provide rigid, pre-defined templates that fail to adapt to a user's evolving biomechanical constraints, while enterprise gym management systems remain operationally isolated from the actual health metrics of their members. To address these limitations, recent research has explored the integration of Large Language Models (LLMs) to synthesize personalized workout and nutritional regimens dynamically. This survey evaluates the state-of-the-art in automated fitness recommendation engines and gym operational frameworks, analyzing the trade-offs between legacy rule-based systems and modern generative APIs. By establishing a taxonomic classification of existing technologies, this review exposes critical gaps in cross-platform orchestration, highlighting the necessity for unified architectures—such as the proposed FitCore paradigm—that seamlessly combine conversational AI with enterprise resource management to maximize user progression and operational efficiency.

A Network for Monitoring and Assessing Water Quality for Drinking and Irrigation Purpose

Authors: Mrs.P.Sandhya Krishna,, Yelisetti Chaitanya, Parimi SwarnaLatha, Kopparapu Pushpa vani,, Divyakolu Ahalya

Abstract: By consistently monitoring this, our drinking and farm water, for bad stuff and acting intelligently on the information we are given, we can keep it safe. This paper is on a system using connected sensors and smart computer programs to monitor water quality. The system consists of a suite of low-cost sensors capable of determining pH, clarity of the water, its temperature, oxygen levels in the water, and how much material is dissolved within it.These sensors send data over wireless to a computer program that organizes and checks the data using smart programs such as Support Vector Machines, Random Forests, and Artificial Neural Networks. These programs help in deciding if the water is good enough for drinking or to be used in crops based on guidelines provided by the World Health Organization and our country. It picks out the most useful data and spots anything unusual to make sure its guesses are right, and can even break above 95% in accuracy. The system can also guess when the water might get bad, so we can fix it early. All in all, this is a system based on sensors, smart data analysis, and programs, which help us make good choices. It is just an energy-efficient, not-too-costly way to keep an eye on water quality. This helps protect the health of the people and keeps our farms productive.

DOI: https://doi.org/10.5281/zenodo.20676112

Early Detection of Anemia Using Supervised Machine Learning Algorithms

Authors: SK.Sharmila, Bandla Manasa, Burugupalli Jahnavi Krishna, Garine Akansha, Gontu Bhavya Reddy

Abstract: Anemia is among the top causes of hematological disorders that have a serious impact on the health of millions of people worldwide and is marked by a shortage of red blood cells or reduction in hemoglobin concentration that leads to oxygen transport to body tissues becoming impaired. Identifying anemia at the earliest stage is extremely important in preventing the development of serious complications, lowering death rate, and enhancing the quality of life of the population especially women and children. Though accurate, the traditional diagnostic methods are quite lengthy, require lots of resources and are difficult to be accessed in places with few resources. This research is about the use of different supervised machine learning algorithms to generate a model that can predict anemia at the initial levels from blood test parameters.Various models such as Decision Tree, Random Forest, Support Vector Machine, Logistic Regression, and K-Nearest Neighbors were trained and evaluated with a labeled dataset comprising clinical and blood test features like hemoglobin level, hematocrit, RBC count, and mean corpuscular volume. The dataset was preprocessed to account for missing entries, normalize scales, and optimize feature importance through correlation analysis and recursive feature elimination. Metrics such as accuracy, precision, recall, F1-score, and ROC-AUC were used for comparing the models' performance. Experimental results suggest that ensemble-based algorithms, especially Random Forest, had better predictive accuracy and interpretability. The results indicate that machine learning is a viable tool for healthcare professionals to detect anemia at an early stage, thus allowing for the provision of appropriate treatment and timely intervention. These results pave the way for the seamless incorporation of AI-driven diagnostic tools into everyday healthcare screening routines.

DOI: https://doi.org/10.5281/zenodo.20676314

Edge-Enabled AI Surveillance Framework for Real-Time Multi-Threat Detection (EAS-RTD)

Authors: Assistant Professor Mrs. G. Rohini Phaneendra Kumari, Desineti NagaLakshmi, Gangavarapu Sailaja,, Yaganti Pavani, Nalluri Gayatri Priya

Abstract: The increasing complexity of urban environments and open spaces calls for sophisticated surveillance systems that can identify threats instantly. Human operator-based traditional video monitoring methods are often inefficient, error-prone, and have limited scalability. This paper introduces a smart surveillance system that integrates deep learning, machine learning, and computer vision technologies to detect the occurrence of weapons, violent acts, fire, or smoke in real time.Video streams are locally processed using edge computing which reduces latency and network congestion and, at the same time, allows devices with limited resources to operate efficiently. Compact convolutional neural networks along with object detection algorithms such as YOLOv8 are used to obtain precise classification and tracking, whereas Explainable AI provides interpretability and human trust in automated decisions.Along with adaptive ML-based security solutions to defend the edge devices from cyber-physical attacks, secure data transmission, camera installation optimization, multi-threat detection, and crowd behavior analysis are some of the other features of the system. The experimental evaluation, as conveyed, is accurate, has low latency, and is scalable, thus, the system is suitable for smart cities, transportation hubs, and critical infrastructure. The paper positions AI-powered, edge-enabled surveillance as a way to improve situational awareness and enhance public safety.

DOI: https://doi.org/10.5281/zenodo.20676682

Prediction of Soil Moisture for Smart Irrigation Using Machine Learning Techniques

Authors: Assistant Professor Mr.K.Srikanth, Dandu Amulya, Tanguturi Kameswari Smruthi, Gonuguntla Triveni,, Myneni DattaSri

Abstract: Water shortages combined with the rising demand for more food production have put pressure on the need for intelligent irrigation management. To address these challenges, this study devised a machine learning–based solution to soil moisture prediction that would enable smart irrigation practices to conserve water and increase crop yield. In order to derive accurate moisture content, the proposed model gathers environmental parameters such as temperature, humidity, rainfall, air pressure, and soil nutrients and then executes one linear regression algorithm. A Django web interface is embedded with the module through which farmers can feed in environmental data and get irrigation advice immediately. The test outcomes confirm that the application attains very high prediction accuracy and can save enormously on water. Thus, by integrating machine learning with web-based automation, a feasible, scalable, and resource-friendly solution is available for agriculture equipped with modern technology, especially for those places that suffer from water shortage.

DOI: https://doi.org/10.5281/zenodo.20677077

Heart Disease Risk Prediction Using Hybrid Machine Learning Approaches

Authors: Assistant Professor Mrs. G. Lakshmi Durga, Lakka Pranavi, Kakarla Veera Venkata Lakshmi Poojitha,, Pagolu Naga Divya,, Chiluvuri Sucharitha

Abstract: Heart disease is one of the main causes of death around the world. This shows the urgent need for effective prediction systems to prevent serious heart events. This study introduces a Hybrid Stacked Machine Learning Model (HSMLM) for predicting heart disease risk. It combines traditional and ensemble-based classifiers to improve diagnostic accuracy and reliability. Logistic Regression, Support Vector Machine (SVM), Random Forest (RF), XGBoost and K-Nearest Neighbour (KNN) are the algorithms used by the framework. These algorithms are used in an optimized stacked ensemble environment. The features are selected by determining their importance with Chi-Square, ANOVA and Mutual Information to get the most important clinical factors. Class imbalance is also addressed to enhance performance with the Synthetic Minority Oversampling Technique (SMOTE) and cross-validation. The modelling in analyzed against well-known standard datasets with UCI and Kaggle heart disease datasets to confirm validity and attain 92% accuracy, 92% F1-score and 94% ROC-AUC, which outperformed individual models by 3-4%. This demonstrated that using a hybrid model provides significant improvement with predictive credibility with reasonable interpretative clarity. It shows progress in approach as an innovative decision-support tool for anticipated early diagnosis support. The HSMLM approach provided opportunities for both practical decision-making and clinical inference with the goal of reducing deaths caused by heart disease while improving patient outcomes.

DOI: https://doi.org/10.5281/zenodo.20677303

Analytical Study On Impact And Effect Of Highway Development On Environment And Communities

Authors: Vandana Sharma, Dr Subha Khatri

Abstract: Although the effects of highway development are similar to those of other human activities that harm the environment, highways (as well as power line rights-of-way and other transportation routes) have special effects because of their linear shape. In forested landscapes, highways function as concave corridors, which are areas with lower vegetation heights than the surrounding habitat matrix; in agricultural and some rangeland landscapes, where dense vegetation is encouraged along the roadsides, highways may function as convex corridors. In addition to improving the quality of current roadways, highway development increases connection between major economic hubs. Highway expansion is a result of growing traffic and the need to support the region's economic potential. Additionally, the accession activity alters the surrounding landscape and disrupts the environment. Additionally, it has both direct and indirect effects on biotic and abiotic components. Therefore, it is essential to conduct an Environment Impact Assessment of National Highways in order to understand and forecast the effects on the environment and the socioeconomic circumstances of the inhabitants. Thus, this study examines how highway growth affects the quality of the air, water, and soil as well as the socioeconomic circumstances and health.

IoT-Based Red Light Violation Detection and License Plate Tracking Using ESP32

Authors: Assistant Professor Mrs.Aruna Kumari, Nimma Himani, Kantipudi Jasmitha,, Korivi Naga Krishnaveni, Kakumanu Manikanteswari

Abstract: Traffic congestion and frequent red-light violations in urban areas require smart, automated monitoring systems. This paper presents Auto-ViTrack, an IoT-based framework for real-time red-light violation detection and license plate tracking. The system uses an ESP32 microcontroller along with infrared (IR) and ultrasonic sensors to detect unauthorized vehicle movement during the red signal phase. When a violation is detected, event data including timestamp and sensor readings is sent to the ThingSpeak cloud for remote monitoring and analysis. A Convolutional Neural Network (CNN) module helps recognize license plates to identify offending vehicles accurately. This hybrid IoT and ML architecture enables event-driven data logging, reduces redundant transmissions, and optimizes bandwidth usage. Experimental results show that Auto-ViTrack outperforms existing systems like Vision-SORT, Smart-Enforce, and IoT-VMS in detection accuracy (94%), recognition reliability (92%), and latency reduction (350 ms). This cost-effective and scalable solution improves traffic law enforcement, enhances road safety, and supports the creation of sustainable smart city infrastructure.

DOI: https://doi.org/10.5281/zenodo.20677741

Loan Eligibility Prediction Using Hybrid Machine Learning Models

Authors: Assistance Professor Mr. R. Srinivas, Kattamuri Lakshmi Gayathri,, Kommalapati Bindu Sri,, Baligodugula Lakshmi Kranthi, Purimetla Jahnavi

Abstract: Loan eligibility prediction is a critical task in the financial sector, enabling institutions to assess applicants’ creditworthiness accurately and reduce default risks. Traditional machine learning models often struggle with heterogeneous and imbalanced financial data, which affects prediction reliability and fairness. This study proposes a Hybrid Machine Learning Framework (HMLF) that integrates feature selection, ensemble learning, and optimization to enhance predictive performance and interpretability. The framework combines Logistic Regression, Random Forest, Gradient Boosting, and Deep Neural Networks within a stacking ensemble to leverage the complementary strengths of each model. Feature engineering, normalization, and Synthetic Minority Oversampling Technique (SMOTE) are applied to improve data quality and class balance. The hybrid model is trained and validated on benchmark financial datasets using cross-validation to ensure generalization. Experimental results show that the proposed approach achieves higher accuracy, precision, recall, and F1-score compared to traditional single-model classifiers. The ensemble design improves stability and reduces bias in decision outcomes. The findings highlight that the proposed hybrid system provides a reliable, transparent, and scalable solution for automated loan eligibility prediction, supporting financial institutions in making data-driven and fair lending decisions.

DOI: https://doi.org/10.5281/zenodo.20677984

IoT-Based Anemia Detection System Using Temperature, Heartbeat, and Oxygen Sensors

Authors: Assistance Professor Y.V.Vamsi Krishna Teja, Bodduluri Deepika, kota Sri Tany, Bavanari Dhanya Sri, Rayudu Kavya Chowdary

Abstract: Anemia is a widespread health condition caused by a deficiency of hemoglobin or red blood cells, leading to fatigue, weakness, and other serious health complications. Early detection and continuous monitoring are crucial for effective treatment and prevention. Traditional diagnostic methods, such as invasive blood tests, are time-consuming, costly, and often inaccessible in rural or resource-limited areas. This study presents an IoT-based anemia detection system (HI-MASS) that utilizes temperature, heartbeat (BPM), and oxygen (SpO₂) sensors to monitor physiological parameters that correlate with anemia. The system collects real-time data through wearable devices, which are then transmitted to cloud servers and processed using predictive algorithms to detect anemia, classify its severity, and provide timely alerts. The integration of IoT technology with machine learning enables non-invasive, cost-effective, and real-time health monitoring, improving accessibility for remote populations. Experimental results demonstrate high reliability and strong potential for continuous anemia monitoring, making the proposed system a scalable and practical solution for proactive healthcare management.

DOI: https://doi.org/10.5281/zenodo.20678355

Development Of A Smart Interview Assistant For Real-Time Candidate Evaluation

Authors: k. Rajkumar, Donthuri Pranitha

Abstract: Automated intelligent software agents may mimic human conversational behaviours, allowing for more natural and interesting interactions with people, thanks to the fast development of Conversational AI. By allowing for the replacement of human interviewers with intelligent autonomous software agents, these developments pave the way for the automation of the candidate interview process. Agents outfitted with Conversational AI can mimic human interviewers in every way: asking questions, comprehending and evaluating responses, and starting up dynamic discussions. More effective and equitable recruiting processes are the result of this automation, which streamlines the interview process as a whole and guarantees consistent and impartial judgement. The purpose of this research article is to provide a thorough analysis of the design and implementation of an AI-driven interview system for real-time applicant assessments. As part of the system, various AI agents carry out various tasks, such as choosing from a set of predefined questions, assessing candidates' responses, analysing speech for sentiment and emotion, and finally, combining these analyses to generate separate scores for answers and emotions in the performance evaluation.

DOI: https://doi.org/10.5281/zenodo.20678426

AI-Based Hardware Encryption And Threat Detection Using Zynq PS-PL Architecture

Authors: Gokularangan V, Narsingam K, Dr. V. Ramesh Kumar

Abstract: This paper presents a comprehensive technical account of an AI-based hardware encryption and threat detection system implemented on the Xilinx Zynq-7000 SoC using the ZedBoard development platform. The design exploits the tightly-coupled Processing System (PS) and Programmable Logic (PL) of the Zynq architecture to deliver real-time security capabilities operating at hardware speed. The core innovation is the fusion of a lightweight AI decision model (threshold-based anomaly detection) with a hardware XOR encryption accelerator, both implemented in the FPGA fabric and exposed to the ARM Cortex-A9 processor through an AXI4-Lite memory-mapped interface. The system achieves sub-nanosecond encryption decisions with less than 0.02% LUT utilization on the xc7z020clg484-1 device. Behavioral simulation confirms correct ciphertext computation and accurate anomaly flagging across five boundary test cases, and bitstream generation completes with over 7 ns of positive setup slack at 100 MHz. The architecture is intentionally modular and parameterizable, providing a solid foundation for upgrading to production-grade AES encryption and hardware neural network-based anomaly detection in future work.

Enhancing Heart Disease Prediction Accuracy Using Hybrid Machine Learning Techniques

Authors: B. Naresh, C. Jyothi

Abstract: Researchers have shown a lot of interest in the field of medical science. A fair amount of researchers have identified many reasons for early death in humans. According to the relevant literature, there are several causes of diseases, and heart-related illnesses are one of them. In an effort to save lives and aid medical professionals in the detection, prevention, and management of cardiovascular disease, several researchers have put forward unique approaches. Every effective plan has its limitations, but there are several easy approaches that help the expert make a conclusion. The two feature selection approaches, Correlation Based Feature Selection (CFS) and Gain Ratio, as well as the Hidden Markov Model (HMM), Artificial Neural Network (ANN), Support Vector Machine (SVM), and Decision Tree J48, are thoroughly examined using the suggested method. When applied to a separate set of data, the Ranker technique is complemented by the Gain Ratio. The proposed technique analyses the process and then intelligently constructs Naive Bayes processing by combining the two best processes using an appropriate layered architecture. Choosing the best approach and comparing the available schemes with various characteristics for statistical analysis is the primary goal at the outset.

DOI: https://doi.org/10.5281/zenodo.20678726

Intelligent Honey Pot-Based Framework For Modern Cyber Security Defense

Authors: Parisa Premchand Goud, Gadde Anusha

Abstract: A comprehensive analysis of honey pot integration with various cyber security frameworks is presented in this paper. These frameworks include firewalls, Intrusion Detection and Prevention Systems (IDPS), Security Information and Event Management (SIEM) systems, and Security Orchestration, Automation, and Response (SOAR) platforms. The paper builds a framework for improving honey pot functions using AI and ML using a systematic research process that includes literature review and case-based analysis. Adaptive responses to sophisticated cyber attacks, predictive analytics, and dynamic threat information may all be generated using this novel technique. The results highlight the importance of honey pots in improving the accuracy of threat detection, decreasing resource overhead, and giving useful information on the strategies, methods, and procedures (TTPs) used by attackers.

DOI: https://doi.org/10.5281/zenodo.20678983

Predicting Agricultural Productivity Using Machine Learning Models

Authors: Mrs. G. Rohini Phaneendra Kumari, Thalla Vasavi Surya Prabha, Bandla Hima Varshini, Kilaru Vyshnav, Domathoti Nadiya

Abstract: One of the most important prerequisites for sustainable agricultural planning, food security, and proper resource management is being able to accurately predict crop yield. The nonlinear and intricate interactions between environmental variables and crop productivity are often overlooked by traditional statistical models based on limited historical data. This work is a machine learning exploration to develop a model that predicts food production given the soil, weather, and crop-related parameters. Our dataset has environmental and agricultural features that include soil nutrients like (N, P, K, pH), atmospheric variables (temperature, relative humidity, and rainfall) and crop-specific attributes. By mixing these factors together the research intends to deliver an intelligent predictive framework that would be very useful for farmers and policymakers in decision-making and agricultural planning. To get better prediction accuracy, various machine learning algorithms including Random Forest (RF), Support Vector Machine (SVM), and XG Boost were applied and compared. The quality of each model was measured by R² score, RMSE, and MAE metrics. Random Forest was the best performer and therefore had the best precision and stability in capturing nonlinear data patterns among the models that were tested. The findings serve as evidence of the potential of machine learning methods to revolutionize the Agri-ecosystem by turning the traditional farm practices into data-driven decision-making, which is a significant contributor to the implementation of sustainable crop management and enhanced yield forecasting.

DOI: https://doi.org/10.5281/zenodo.20679086

Secure Digital Voting System Using Blockchain And Web Technologies

Authors: S. Venkateswara Rao, Kota Spandana

Abstract: In order to solve the problems with conventional offline voting in India, which include expensive prices, staff needs, delayed results, and accessibility concerns, especially for NonResident Indians (NRIs) and users with technical knowledge, we need to create a safe and effective online voting system. First, there's the Authentication Phase, when users are checked using Face Recognition and OTP Verification. Then, to make sure everything is honest, there's Live Real-time Monitoring and Session Monitoring. If anything fishy is found, the vote won't go through. Problems arise for Non-Resident Indians (NRIs) because to the high expenses and substantial human effort needed by India's current voting system, which uses Electronic Voting Machines (EVMs) and conventional ballot papers. To guarantee the safety of online voting, the suggested system incorporates several security measures, such as Aadhaar-based authentication with OTP, MTCNN for face recognition, MobileNetV2 for face real-time monitoring, Blockchain-based smart contracts for secure verification, and end-to-end encryption. Due to logistical constraints, the participation of non-resident Indian voters in conventional elections has been low. Because of the increased accessibility and convenience offered by an online voting system, projections indicate that participation will increase by 5% in 2024, reaching 72%. A more inclusive voting process and more participation are outcomes of this digital transition, which removes geographical obstacles (p = 0.049). By using cutting-edge technology like as TensorFlow and MobileNetV2, the created system successfully overcomes the shortcomings of conventional offline voting. Aligning with the changing demands of future generations and offering an inclusive solution for all voters, it provides the groundwork for a democratic voting process that is entirely digital, safe, and efficient.

DOI: https://doi.org/10.5281/zenodo.20679200

Optimizing Real-Time Sign Language Detection Using Deep Learning and Computer Vision

Authors: Assistant Professor MR.G.K.Harinadh, Aneesa Shaik, Nuthi Vyshnavi,, Kurra Srivalli,, Madiraju Bhargavi

Abstract: Sign language is an essential way for people with hearing and speech impairments to communicate. However, for those who don’t know it, understanding sign language can be difficult.This paper introduces a real-time Indian Sign Language (ISL) recognition and translation system that uses deep learning and computer vision techniques. The application captures video from a webcam, identifies ISL gestures, and translates them into a spoken language like English, allowing for easy interaction.To accurately interpret hand movements, the system uses a Convolutional Neural Network (CNN) model trained on a tailored ISL dataset. A thorough preprocessing pipeline, which includes background removal, contour detection, and image normalization, improves the model’s performance under varying lighting and environmental conditions.

DOI: https://doi.org/10.5281/zenodo.20679330

Machine Learning-Based Insider Threat Detection For Enhanced Organizational Security

Authors: Tuniki Pravalika, Panuganti Lavanya

Abstract: The potential compromise of sensitive information and company assets by employees is a major concern for any firm. Robust ML algorithms that can handle complicated and biassed data are necessary for the threat detection process. Some of the ML models that are tested in this study using the well-known CERT dataset include Logistic Regression, Decision Trees, Random Forest, SVM, KNN, Naïve Bayes, Adaboost, and XGBoost. Approaches like SMOTE, which deal with problems brought on by data imbalance, emphasise the need of a balanced dataset. A 97.5% success rate in detecting insider threats was achieved using Random Forest and Adaboost, according to the data. This study lays the groundwork for more trustworthy organisational security measures by improving approaches for identifying insider threats and offering a systematic evaluation of model performance. A few of the terms that come up include SMOTE, CERT, insider threat detection, and machine learning.

DOI: https://doi.org/10.5281/zenodo.20679357

Integration of IoT Sensors and Cloud Analytics for Intelligent Threat Response

Authors: Associate Professor K.Jagadeesh, Betapudi Gagana, Kota Chekitha Nagalakshmi, Beeraka Yasaswi Subhagatri, Kunduru Keerthi

Abstract: The number of Internet of Things (IoT) devices has exponentially increased the amount of fresh data from real-time sources in various fields such as smart cities, healthcare, manufacturing, and critical infrastructure. Nevertheless, the problem of handling and interpreting the huge volume of diverse data produced by these devices still stands, especially when it comes to promptly identifying threats and reacting to them. This paper presents a merged IoT-cloud threat response system that interconnects the heterogeneous IoT sensors with the cloud-based processing accompanied by machine learning techniques to sense the irregularities and foresee the security risks even before they emerge. At the same time, edge computing is cutting down on the delay and improving the responsiveness. Notably, the system is overcoming the issues of infinite scalability, data diversification, communication lags, and security loopholes via the effective handling of data, the initializing of self-adjusting learning models, and the use of the safe transmission protocols. The experimental results confirm the system's superiority in terms of detection accuracy, response time, and operational efficiency over the competitive methods. As the discussed scheme makes threat management more proactive and predictive, it can be considered as a versatile and scalable next-gen solution for smart surveillance and cybersecurity scenarios.

DOI: https://doi.org/10.5281/zenodo.20679501

IoT-Based Smart Railway Gate Control System for Enhanced Safety and Automation

Authors: Mr. G. K. Harinadh, Ganga Tanmai, Yemineni Yamini Kalayani, Pujita Dondapati

Abstract: The document gives the details about the creation of an automated and smart railway crossing system with an IoT-based controller. The project goal was to increase safety and save time at railway crossings, mostly in those areas where the gates are still manually operated. It has been found that manually-operated systems are slow in their reaction, over-rely on human vigilance, and often waste time and energy in the course of their ordinary operation. Some so- called "automated" solutions hardly rely on the use of cameras or a single type of sensors to locate a train that is coming. Even if these technologies were to function in perfect conditions, they would still have problems in real cases. The cost of their installation is oftentimes not very affordable, the range of detection can be quite small, and such factors as bad weather, darkness, and dust can deteriorate the performance of devices. Our setup solves that with the help of the NodeMCU microcontroller which acts as the central controller and connects all the parts. Instead of relying on only one source of detection, several sensors are employed to increase reliability. Reed sensors are placed to detect the arrival of a train, infrared sensors are used to detect if there are any objects near the gate or on the track, and a geared motor mechanism is used for smooth gate movement. Supplementary safety features have also been added so that the operator will instantly be notified if something goes wrong. The system has the advantage of IoT which allows it to connect to the Internet. As the system is able to wirelessly communicate with the world outside, it is possible to track the status of the gate from any location, and a malfunction can be identified almost immediately. This will allow the manual checking to be done at intervals and not all the time, and also the system will be less troublesome to maintain. In our research, the setup we propose was found to have a better performance than a number of traditional ways of railroad monitoring such as TinyML-based tracking and GPS-enabled systems. The usage of various sensor types along with continuous data processing and wireless communication leads to much more accurate and speedy gate operations. As a consequence, the total duration of gate closure is shortened which means fewer vehicles are forced to wait and there is a lower risk of accidents. In brief, the presented system is a cost-saving, expandable, and intelligent IoT-based solution to the problem of railway crossing safety. It is in line with the trend of intelligent transportation systems and serves as an example of how current IoT technologies can be used to increase the reliability of the most vital public infrastructure.

DOI: https://doi.org/10.5281/zenodo.20679708

A Review On Toxic Effects Of Fungicide On Various Physiological Aspects Of Fish

Authors: Tanya, Neeru Singh

Abstract: Ecotoxicology is the science of the effects of toxic chemicals on biological organisms, especially at the population, community, ecosystem, and biosphere levels. Pesticides, after surface runoff, enter the aquatic system, causing a hazardous condition in the environment. These agrochemicals have been reported to be highly toxic not only to the fish population but to the other organisms as well, which form the food chain of the ecosystem. The contaminations of water result in various adverse effects on the health of the fish by altering their physiological status. The fungicides cause toxic effects on the biochemical, haematological, genetic, enzymatic, protein, and histological status of the organisms. In the course of this review, an attempt has been made to explain the ill effects of the commonly used fungicides on the different species of fish by rating the toxicity levels of the toxicants. fish metabolism is directly affected by different pesticides. They cause different types of toxicity in fish such as changes in fish behaviour, haematological changes, histopathological disturbances, biochemical modifications, endocrine system disruption, changes in antioxidant defence system and alteration of acetylcholinesterase activity. The judicious use of fungicides must be carried out by the farmers. The use of Biopesticides and natural products must be encouraged to avoid the toxic effect of the fungicides. Biocontrol of the pests must be a wise selection as compared to the use of synthetic agrochemicals. The alternations or changes that this review will cover include the following: alternations or changes that affect the behavior of the fish, such as increased breathing on the surface, tremors on the fins of the fish, gulping of air, irregular swimming, and movement of the operculum; alternations or changes that affect the biochemical characteristics of the fish, such as the damage to the antioxidant defense system of the fish due to the formation of ROS, alterations in the levels of various indicators of oxidative stress.

DOI: http://doi.org/10.5281/zenodo.20681485

SafeDrop: QR Enabled Delivery Bot

Authors: Ms. Mansi Sunil Patil, Ms. Sakshi Yashwant Kalkute, Ms. Jayshree Dnyaneshwar Wagh, Mrs. Sarojini Naik

Abstract: As a new Application to our everyday life. We designed and implemented the SafeDrop: QR-enabled delivery bot, which is an innovative automation system designed to increase efficiency and quality of the services. It is a cost-effective, and scalable solution for automating indoor package delivery. The system utilizes the ESP32 for communication tasks, sensing and to control navigation. It integrates autonomous navigation using IR sensors, secure QR- based authentication and multi-compartment design. Manual delivery of parcels in environments such as restaurants, hospitals, and offices demands continuous human involvement. often resulting in delivery delays, inefficiency, human errors, and compromised security. IR sensors are employed for obstacle detection and line following and ensuring accurate movement, while DC motors and a motor driver enables efficient mobility. A QR-based authentication mechanism enhances security by allowing access only to the authorized users. The delivery box is opened by servo motors and the OLED screen shows the status of the bot. The proposed solution is a cost-effective and reliable way for contactless delivery, and demonstrates the practical implementation of IoT and embedded technologies in real-world applications.

Formulation And Evaluation Of Herbal Soap Using Medicinal Plant

Authors: Mr. Darshan Rajput, Mr.Pranay Kamble, Ms.Pratibha Makar, Dr.Vijaykumar Kale, Dr.Mahesh Thakre

Abstract: Herbal soap is a natural skin-care product made from plant-based ingredients such as neem, tulsi, turmeric, aloe vera, tomato, beetroot, apricot, coconut oil, eucalyptus oil, vitamin E, and glycerin. Unlike many commercial soaps that contain synthetic chemicals, herbal soap uses natural materials that are gentle on the skin. It helps cleanse the skin without causing dryness or irritation. The soap was prepared by melting the glycerin base, mixing it with herbal extracts and oils, and then pouring the mixture into molds for cooling and solidification. The finished soap was dark brown in color, had a pleasant mild fragrance, and produced a satisfactory amount of foam. Its pH ranged from 7.0 to 7.3, making it suitable for regular skin use. The soap showed good stability, retained moisture, and did not cause redness or discomfort. Overall, it is a safe, eco-friendly, and effective herbal cleansing product.

AI-Enabled Critical Care: Advancing Diagnosis, Triage, and Treatment Decision-Making

Authors: Oyeleke Stevens O, Adebayo Adekunle A, Oladokun David O, Ikotun Olufunmilayo, Dr Oyeleke S.O

Abstract: Background: Intensive care units generate high-frequency, multimodal data suited to artificial intelligence (AI)-enabled clinical decision support systems (CDSS), yet clinical translation remains inconsistent. Objectives: To synthesise evidence on AI-CDSS in adult critical care, appraise methodological maturity against AI reporting standards, and identify implementation barriers. Methods: Narrative synthesis following the SANRA framework. Literature was appraised against TRIPOD+AI, PROBAST+AI, DECIDE-AI, and CONSORT-AI standards across six dimensions: validation tier, data generalisability, metric completeness, human-AI workflow integration, equity reporting, and deployment status. Results: Diagnostic AI models report high internal discrimination but inconsistent calibration and limited prospective validation. Triage systems outperform static scoring retrospectively, yet alert fatigue and clinician override rates remain underreported. Treatment decision support models often conflate observational prediction with causal intervention effects. Cross-cutting gaps include single-centre training data, heterogeneous equity reporting, and absence of standardised post-deployment monitoring. Conclusions: AI-CDSS in critical care exhibits strong algorithmic promise but fragmented clinical validation. Priorities include prospective multi-centre evaluation, human-centred workflow integration, causal treatment framing, and standards-compliant reporting.

DOI: http://doi.org/10.5281/zenodo.20693087

KumbhSahyogi A Service Provider Platform: Smart Transport And Parking Management System

Authors: Vaishnavi Chaudhari, Priti Lahane, Mayuri Daga, Prathamesh Patil, Kailas Adhav

Abstract: Maha Kumbh 2027 at Nashik is expected to attract millions of devotees from around the world. This presents significant challenges in managing transportation, parking, and emergency response. Traditional systems often fail to handle large crowds effectively, resulting in congestion, delays, and security problems. This paper introduces KumbhSahyogi, a GPS-based, AI-powered smart mobility and service management system de- signed to maximize parking allocation, optimize routes, and coor- dinate transportation for large events like the Kumbh Mela. The system includes various features such as real-time parking slot availability, AI-powered route optimization through Dijkstra’s algorithm and congestion heatmaps, multimodal transport inte- gration, and a smart chatbot for user support. The backend uses Node.js and Express along with MongoDB Atlas to manage data securely and effectively. The system greatly improves safety, reduces traffic congestion, and increases user convenience through smart routing, predictive analytics, and instant SOS communication. Experimental testing under simulated event conditions shows faster response times and optimized crowd movement, demonstrating the platform’s potential to efficiently manage large-scale religious and public events.

DOI: http://doi.org/10.5281/zenodo.20698581

Adaptive BNPL With AI: A Comprehensive Study

Authors: Shrutika Giri, Gayatri Badgujar, Anvi Kore, Ravindra Apare

Abstract: Traditional Buy Now, Pay Later (BNPL) services are held back by their rigid, outdated systems, making them easy targets for fraud and unable to offer personalized experiences. This new Adaptive BNPL with AI is designed to overcome these limitations. Solution is a smart, multi-layered system that uses artificial intelligence to boost security and improve the user experience. At its core, it relies on AI models like Logistic Regression and Random Forest to create dynamic credit scores and accurately predict potential late payments. We're also using Isolation Forest for proactive fraud detection, stopping fraudulent activity before it even happens.To make the service even more user-friendly, we're integrating a conversational AI assistant powered by Natural Language Processing (NLP). The system will also feature flexible payment options, including the ability to split payments for a single transaction across multiple items or even multiple people.

DOI: http://doi.org/10.5281/zenodo.20699043

Real-Time Inventory Prediction Using IoT And Time-Series Machine Learning

Authors: Dr. Pankaj Malik, Ashwarya Garg, Akshat Carpenter, Aditya Verma, Alfaiz Mansoori, Vinayak Mishra

Abstract: Real-time inventory prediction is a major challenge in modern supply chain and warehouse management systems due to fluctuating customer demand, delayed stock updates, and inefficient forecasting methods. This research proposes an intelligent IoT-enabled inventory prediction framework using Time-Series Machine Learning techniques to improve inventory visibility and forecasting accuracy. The proposed system integrates IoT devices such as RFID tags, smart shelves, barcode scanners, and environmental sensors to continuously collect real-time stock movement and warehouse data. The collected streaming data is processed through preprocessing and feature engineering techniques before applying forecasting models including ARIMA, Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Prophet, and XGBoost. The system predicts future inventory levels, reorder points, and demand fluctuations in real time, enabling proactive inventory management and automated replenishment decisions. Experimental evaluation was conducted using retail inventory datasets and simulated IoT warehouse data. The performance of the models was evaluated using Mean Absolute Error (MAE), Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), and prediction accuracy metrics. Results indicate that the proposed LSTM-based IoT inventory prediction model achieved the best forecasting performance with an accuracy of 96.2%, RMSE of 8.14, and MAPE of 4.8%, outperforming traditional ARIMA and statistical forecasting approaches. The proposed framework reduced stockout situations by 28% and excess inventory costs by 21% compared to conventional inventory management systems. The integration of IoT and Time-Series Machine Learning significantly improved real-time inventory monitoring, warehouse efficiency, and supply chain responsiveness. The proposed research contributes toward intelligent Industry 4.0-based smart warehouse and predictive supply chain management systems.

DOI: http://doi.org/10.5281/zenodo.20700236

Women Safety System Using IoT-Based Wearable Device With GPS And GSM Integration

Authors: Abhishek Kumar Pandey, Anshul Sahu, Harshit Patwari, Kamlesh Kumar Mishra, Manish, Pankaj Sharma

Abstract: Women safety has emerged as a major concern. This paper presents an IoT-based wearable safety system using Arduino Uno, GPS, GSM, NodeMCU and multiple sensors for automatic emergency detection and alert generation. The system continuously monitors user conditions and sends location-enabled emergency notifications. This paper presents a novel wearable device integrating multiple advanced sensors for woman safety.

DOI: http://doi.org/10.5281/zenodo.20700788

AI-Powered Personalized Storybook Generator – Using Multi-Modal Content Generation

Authors: Anushka Satav, Siddhi Shinde, Ashish Singh, Supriya Jagtap

Abstract: This paper showcases an AI-driven system that creates and shares beautifully illustrated kids’ storybooks via a seamless, all-in-one multimodal workflow. Kick off with a basic user prompt and chapter count, and a large language model (LLM) delivers a ready-to-go story: a snappy title, cover idea, and a neat lineup of chapters each packed with engaging text and tailored illustration prompts. By crafting the whole multi-chapter arc in one shot, it locks in smooth, consistent storytelling from start to finish. Spot-on images pop up for every chapter, pulled straight from the scene descriptions and woven into an interactive reader. Text-to-speech (TTS) kicks in too, voicing the tale aloud to make it more accessible and captivating. The prototype doesn’t yet handle story carryover across sessions or live text syncing with audio. It also skips built-in safety checks, pointing to age-filtering as prime territory for upgrades. All told, this setup proves a smart, streamlined way to automate rich, multimedia stories.

DOI: http://doi.org/

AI-Augmented Reflective Journaling: Transforming Narrative Life-Scripts into Actionable Task Workflows using Large Language Models

Authors: Shreya Parkar, Gayathri Nair, Dr. Jasbir Kaur, Assistant Professor Ifrah Kampoo

Abstract: Traditional personal journaling provides a critical medium for cognitive offloading, yet digital alternatives remain constrained by a passive data lifecycle where unstructured entries reside in isolation from an individual’s productivity workflows. This paper presents LifeScript, a system designed to map qualitative retrospective text into prioritized tasks via a multi-tiered natural language processing framework. The architecture integrates a fine-tuned transformer for contex-tual emotion classification with a neural token classification pipeline to extract behavioral intents, goals, and habits. Ad-ditionally, a heuristic parsing engine normalizes colloquial temporal markers into structured task metrics within a cross-platform client dashboard. Empirical evaluation demonstrates that the framework achieves an entity-level F1-score of 86.3% in intent extraction alongside a System Usability Scale score of 84.5 6.2. These outcomes demonstrate the practical viability of operationalizing natural prose diaries into execution-focused data tracking pipelines, establishing a resource-efficient architecture for personal knowledge management systems.

DOI: https://doi.org/10.5281/zenodo.20813252

MINDFUL: An AI-Powered Mental Health Support Platform For Students

Authors: Nandini Chourasiya, Piyush Ghoshi, Yash Jhanwar, Sayed Ozair Mehmood, Prof. Shahida Khan, Prof. Pawan K. Gupta

Abstract: The escalating mental health crisis among university students, characterized by rising rates of anxiety, depression, academic burnout, and social isolation, demands innovative and accessible support systems. This paper presents MINDFUL, a comprehensive AI-powered mental health and wellness platform purpose-built for students. MINDFUL integrates an intelligent 24/7 chatbot for emotional first aid, a confidential appointment booking system with professional counselors, psychoeducational resources, mood journaling with pattern analytics, crisis support modules, gamified wellness tracking, and a safe peer community forum into one unified ecosystem. The platform employs Natural Language Processing (NLP) for empathetic conversational support, machine learning for emotional pattern detection and burnout prediction, and a React.js and Node.js-based microservice architecture backed by MongoDB for scalable data management. MINDFUL is designed to reduce stigma, improve accessibility, and enable early intervention before mental health issues become severe. Evaluation of the proposed system demonstrates measurable improvements in student engagement with mental health services, reduced response latency for crisis detection, and a significant reduction in counselor appointment scheduling overhead. This paper outlines the system architecture, core modules, technology stack, use case scenarios, and the projected societal impact of MINDFUL on student mental well-being.

DOI: http://doi.org/10.5281/zenodo.20702135

Edu Bridge AI: A Context-Aware Academic Retrieval and Guidance Framework for Intelligent Learning Support

Authors: Mr. Karthik Somasundaram Konar, Dr. Jasbir Kaur, Assistant Professor Ms. Sandhya Thakkar

Abstract: The rapid growth of digital education has significantly transformed the learning landscape. However, This paper presents Edu Bridge AI, an AI-driven academic learning platform built upon a Context-Aware Academic Retrieval and Guidance Framework (CARGF). The framework combines retrieval-augmented academic assistance, intelligent resource organization, personalized study scheduling, and collaborative learning support within a unified educational ecosystem. Unlike conventional learning platforms that provide fragmented educational services, the proposed framework enables context-aware retrieval of academic resources and real-time academic guidance through an integrated AI assistant. Experimental evaluation conducted on a prototype deployment involving 50 student users demonstrated improvements in resource accessibility, user engagement, study planning efficiency, and collaborative participation. The findings suggest that integrating intelligent retrieval mechanisms with educational resource management can enhance learning efficiency and student engagement in higher education environments.

DOI: https://doi.org/10.5281/zenodo.20702322

The SME Reality Gap: Re-evaluating Project Management Standards For Adaptive Framework Development

Authors: Abdulrahman Al-Oksh

Abstract: Small and Medium Enterprises (SMEs) and micro-enterprises form the economic backbone of the global construction sector, accounting for over 97% of active industrial stakeholders and generating more than half of the industry's total value-added output. Despite this overwhelming demographic and economic presence, foundational project management (PM) literature and professional bodies of knowledge remain calibrated for environments of corporate abundance — a condition this study terms the "Academic-Industrial Orientation" (AIO). This study executes a rule-governed Directed Qualitative Content Analysis (DQCA) across a corpus of 769 coded meaning units drawn from Harold Kerzner's Project Management: A Systems Approach to Planning, Scheduling, and Controlling (12th ed.) and the Project Management Institute's A Guide to the Project Management Body of Knowledge (PMBOK® Guide) (8th ed., 2025). The systematic audit empirically confirms a severe "Large-Firm Bias," with indicators of Enterprise-Scale Orientation (ESO) capturing between 68% and 84% of coded distributions across all operational tiers. Drawing on Institutional Theory and Resource Dependence Theory, this paper deconstructs how mainstream frameworks implicitly mandate specialized role segregation, heavy transaction-cost mitigation, and zero-data-friction administrative overhead — bureaucratic burdens that smaller firms simply cannot sustain financially or structurally. This structural mismatch produces what the study calls the "Newly Graduated Engineer's Paradox," trapping entry-level practitioners in a hazardous "Task-Execution Trap" where textbook strategies actively disrupt site velocity and exhaust finite cognitive capacity. To close this reality gap, the paper introduces an SME-Adaptive Engineering Management Framework, featuring a lightweight Project Tailoring Checklist, an administrative decoupling protocol, and an interactive friction calculator. A dual-engine Competency Framework addresses early-career identity shock by reorienting professional capabilities around look-ahead production velocity, relational network governance, and Applied AI autonomy. These tools demonstrate that while standard project management administration is rigidly scale-dependent, core engineering management mathematics remain fundamentally scale-neutral.

DOI: http://doi.org/10.5281/zenodo.20702801

Food Panda Report in Tableau

Authors: B. Sandeep, N. Sathwik, K. Raghavendra, Mrs B Sowjanya

Abstract: Online food delivery platforms have become an essential part of daily life by making food ordering faster, easier, and more convenient. Foodpanda, one of the leading food delivery platforms under Delivery Hero, connects customers with restaurants and delivery partners through a digital platform. This project focuses on analyzing Foodpanda data to understand customer ordering patterns, spending behavior, delivery performance, and regional trends across different locations. The dataset includes information such as customer details, order frequency, restaurant preferences, payment methods, delivery status, ratings, and revenue generated from orders. By analyzing this data through various visualizations and statistical techniques, the project identifies important trends related to customer satisfaction, popular food choices, and business performance. It also highlights how different regions contribute to overall orders and revenue. The findings of this study provide useful insights into customer behavior and operational efficiency. These insights can help food delivery companies improve their services, enhance customer experience, optimize delivery operations, and develop better marketing strategies. Overall, the project demonstrates the importance of data analytics in understanding business performance and supporting informed decision-making in the online food delivery industry.

Manufacturing in the Void

Authors: Samvruth Chamarthi

Abstract: This paper investigates whether the microgravity environment of low Earth orbit provides commercially exploitable manufacturing advantages over terrestrial processes for three high-value material systems: protein crystal growth for pharmaceutical drug discovery, ZBLAN heavy-metal fluoride glass optical fibre, and pharmaceutical polymorph engineering. The study synthesises four decades of peer-reviewed experimental literature alongside verified pricing and mission data from the 2024 commercial milestones achieved by Varda Space Industries and Flawless Photonics. The principal finding is that the suppression of buoyancy-driven convection and gravitational sedimentation in orbit produces measurably superior crystal quality and fibre microstructure for the processes examined, and that this advantage already satisfies the commercial viability condition at current Falcon 9 launch economics for pharmaceutical biologics and specialty ZBLAN photonic fibre.

DOI: https://doi.org/10.5281/zenodo.20711014

Twin-Aware Adaptive Face Recognition Framework Using ArcFace Embeddings And Deep Learning

Authors: Mr. Rohit Sawant, Mr. Rahul Sawant, Dr Jasbir Kaur, Ms. Sandhya Thakkar, Mr. Suraj Kanal

Abstract: Identical twin recognition remains one of the most challenging open problems in biometric security because monozygotic twins share near-identical facial geometry, and ArcFace embedding distributions overlap within the acceptance threshold range (cosine similarity 0.70–0.85), causing conventional systems to fail. This paper proposes a Twin-Aware Adaptive Face Recognition Framework with four novel contributions: (1) an Adaptive Threshold Calibration Engine using T = μ + kσ, derived from intra-class and inter-class embedding distributions; (2) a Twin-Specific Similarity Distribution Analysis module for genetically similar subjects; (3) a structured environmental robustness evaluation with quantified per-condition mitigation strategies; and (4) a real-time ONNX Runtime deployment achieving 28 ms GPU inference latency. Experiments on 10 identical twin pairs (4,000 images, 6 conditions) demonstrate 94.6% accuracy, FAR 4.2%, FRR 5.1%, and AUC 0.96 — outperforming LBPH (78.4%), FaceNet (88.7%), and DeepFace (90.1%). The optimal k = 1.00 was validated by ROC analysis, achieving EER of 4.6%.

DOI: http://doi.org/10.5281/zenodo.20712311

AI-Driven Credit Scoring In Hybrid Cloud Banking: An Investigative Assessment Of Explainability And Regulatory Transparency Gaps

Authors: Amit Butail, Anamika

Abstract: The introduction of Artificial Intelligence (AI) into the credit scoring process has transformed banks' capacity to evaluate risk, however, the deployment of AI models in a hybrid cloud environment brings significant challenges of transparency and explainability in terms of regulations. This study examines the current state of AI credit scoring systems in hybrid cloud banking systems and systematically pinpoints the regulatory transparency challenges that hinder compliance with the banking regulations in the United States, including the Federal Reserve's Model Risk Management guidance (SR 11-7), the Federal Housing Administration's Fair Housing Act (FHA), the Equal Credit Opportunity Act (ECOA), and the National Institute of Standards and Technology's AI Risk Management Framework. To investigate the operational architecture, explainability challenges, and compliance issues of hybrid cloud AI credit scoring systems, a systematic literature review was performed based on the PRISMA framework, which included peer-reviewed publications, regulatory documents, and industry reports from 2020 to 2025. Here are three areas where we found insufficient transparency: (i) the “black box” issue with tracing back opaque decision-making processes due to Federal Reserve Board SR 11-7 remains unaddressed, (ii) lack of explainability mechanisms originating from consumer protection frameworks under the FHA and ECOA, and (iii) lack of transparency and accountability frameworks with AI models in distributed hybrid cloud systems. Therefore, the research contributes to a comprehensive classification of transparency and explainability gaps in hybrid cloud credit scoring, particularly in relation to US regulatory frameworks, and provides basic building blocks for explainable AI (XAI) systems that comply with regulatory requirements.

Bridging The Intelligence Gap: Identifying GenAI And XAI Deficiencies In Banking Risk Management And Regulatory Compliance

Authors: Anamika, Amit Butail

Abstract: The banking sector’s embrace of Generative Artificial Intelligence (GenAI) and Explainable Artificial Intelligence (XAI) is likely to change how financial services firms manage risk, interpret the impact of regulation on their business, and make decisions. Although there is clear commercial interest in both technologies, academic literature has paid comparatively little attention to providing a factual account of their capabilities and limitations. This paper aims to address that gap. Using a systematic review of the literature, practitioner reports, and regulatory guidance published between 2020 and 2025, we identify existing gaps in GenAI technology as applied to credit risk, market risk, operational risk, and anti-money laundering. We also identify existing gaps in XAI techniques, ranging from SHAP-based feature attribution to counterfactual explanation. GenAI technologies present a low-to-moderate risk of hallucination, a lack of transparency in training data, and limited explainability. XAI techniques help but do not fully resolve the explainability problem, particularly in high-frequency trading environments and complex, multi-agent regulatory contexts. We conclude by proposing a pragmatic research agenda for banking technology, risk, and regulatory professionals.

DOI: http://doi.org/10.5281/zenodo.20717108

Review Of Ground-Borne Vibration Mitigation In High-Speed Rail Systems Through Various Modifications

Authors: Gaurav R. Khairnar, Shailesh P. Palekar

Abstract: High-speed rail systems generate significant ground-borne vibrations that can lead to structural damage, discomfort to nearby residents, and accelerated deterioration of railway components. This review investigates various vibration mitigation strategies used in high-speed rail infrastructure, focusing on track and material modifications. Key techniques include the application of resilient rail pads, ballast mats, floating slab tracks, and subgrade treatments. Each method's effectiveness is analyzed based on experimental findings, numerical simulations, and field data. Among these, resilient rail pads are highlighted for their practicality and ease of implementation, offering effective vibration isolation without extensive structural changes. However, the success of mitigation efforts depends on factors such as train speed, soil conditions, and system design. The paper concludes looking at what we still don't know and suggests using several different solutions together to get better results. This review aims to guide engineers and researchers in developing efficient, site-specific vibration control solutions for high-speed rail systems.

DOI: http://doi.org/10.5281/zenodo.20722338

Review On Automated Solar Management System

Authors: Mr V.K.Sambhar, Dr. H.M.Jadhav, Harshal Chaudhari, Shrikrushna Sawant, Akash Dharmadhikari

Abstract: Although solar energy is considered a premier reusable power network for sustained development, it is frequently limited by panel maintenance and energy management of using solar energy and its components. This proposed study demonstrates an Automated Solar Management System consisting of power source management, real-time Monitoring and cleaning in support of increasing usable energy, reliability and operating ability. The solar management system uses an Arduino microcontroller with an MPPT charger, a WIFI module and a motor driver to autonomously commutate power supplies between solar energy, battery backup and grid power supply while yielding energy accessible at all times. This cleaning system relies on a DC motor and WIFI instructions and is a key feature because dust can cause considerable electricity loss up to 30% of its total output. This system will also utilize PWM battery charging features which further increase power harvesting and efficiency in the storage mode to minimize losses achieving the best solar orientation toward the sun by incorporating solar elements of monocrystalline panels and lithium- ion batteries. It provides deployable and low-cost autonomous functions through solar management options, which contributes additional value to the renewable based systems efficiency, and introduces continuous operational functions.

DOI: https://doi.org/10.5281/zenodo.20727095

AI Based Healthcare Chatbot

Authors: Minal Barhate, Pranjal D. Nagmule, Sanskar M. Nalegaonkar, Pranav V. Nagur, Parth D. Nakti, Naaz K. Tadavi

Abstract: The convergence of artificial intelligence (AI) and healthcare presents enormous potential in transforming medical services, diagnostics, and information dissemination. However, a significant challenge remains in ensuring the trustworthiness and credibility of AI-generated information. Many existing AI-powered chatbots rely on large language models (LLMs) trained on diverse and non-specialized data sources, which can result in hallucinations responses that are syntactically correct but factually incorrect. This poses substantial risks in medical contexts, where accuracy is critical. To address this issue, we introduce "AD Bot," a source-grounded medical chatbot designed to generate verifiable responses strictly based on trusted medical literature. By integrating retrieval-augmented generation (RAG) with a vector-embedded knowledge base derived from medical PDFs and textbooks, AD Bot ensures reliable and evidence-based outputs. The system's architecture encompasses PDF text extraction, chunking and embedding, similarity search via vector databases, and interaction with a Hugging Face-hosted LLM. Built using Python, LangChain, Streamlit, and vector search tools such as Pinecone and Chroma, AD Bot offers a secure, modular, and user-friendly platform. This paper details the design, development, and testing of the chatbot while highlighting its implications for the future of AI-assisted healthcare.

DOI: http://doi.org/10.5281/zenodo.20728525

AI Powered Urban Mobility Optimizer With Carbon Footprint Reduction And Smart Carpooling Integration

Authors: Jayashri Waman, Pratik Pawar, Shruti Dalvi, Rohan Mane, Anita Mahajan

Abstract: Urban mobility systems are facing increasing pressure due to traffic congestion, rising fuel consumption, and growing environmental concerns. The transportation sector alone contributes a significant share of global greenhouse gas emissions, making sustainable mobility solutions a critical requirement [1]. Over the past decade, researchers have proposed several optimization-based routing models, including the Pollution Routing Problem (PRP) [5], fuel-consumption-aware vehicle routing approaches [4], and emission-sensitive routing frameworks [3]. However, the majority of these efforts have focused primarily on freight and logistics operations rather than passenger transportation. Advances in time-dependent routing models [10] and environmentally conscious vehicle routing formulations [6], [7] have improved operational efficiency and emission control in logistics networks. Nevertheless, passenger-oriented solutions such as smart carpooling and ridesharing remain relatively underexplored from a carbon-optimization perspective, despite their potential to significantly reduce vehicle usage in urban environments. To address this gap, this paper proposes AUMO, an AI-powered Urban Mobility Optimizer designed to integrate carbon-aware routing with smart carpooling for passenger mobility. The proposed framework adapts established low-carbon routing principles from freight transportation [3], [6] and applies them to urban passenger travel scenarios. A simulation-based evaluation using a microscopic traffic environment demonstrates that the proposed system achieves notable reductions in CO₂ emissions and fleet size while improving vehicle occupancy and overall travel efficiency. These results highlight the potential of intelligent carpooling systems for sustainable urban mobility.

DOI: http://doi.org/10.5281/zenodo.20728800

Heart Disease Risk Assessment Using Machine Learning

Authors: Kiran More, Pranav Upare, Tejas Terdale, Raj Sonune, Pushkar Patankar, Yash Shinde

Abstract: Cardiovascular disease continues to be a major cause of death globally, emphasizing the importance of timely and accurate risk evaluation. This project creates a system based on machine learning to estimate a person’s risk of heart disease by utilizing clinical and lifestyle factors. Numerous supervised algorithms such as Logistic Regression, K-Nearest Neighbors (KNN), Random Forest, and XGBoost—were trained and assessed on a publicly accessible cardiovascular dataset. Following preprocessing and model fine-tuning, performance was evaluated using Accuracy, Precision, Recall, F1-score, and ROC- AUC. Of the models evaluated, Random Forest demonstrated the best predictive capability.

DOI: http://doi.org/10.5281/zenodo.20729166

Autonomous Semiconductor Design Using Reinforcement Learning: Toward AI-Native Chip Engineering

Authors: Arnav Butail

Abstract: The rapid advancement of integrated circuits has led to designs containing hundreds of billions of transistors. Additional months of effort are required for physical design even with advanced Automated Design Tools. This paper assesses the gap closing potential of reinforcement learning, especially as implemented by DeepMind’s proposed AlphaChip, and presents the theory of RL-based design, analyzes the architecture of AlphaChip and Edge-based Graph Neural Networks, evaluates deployment results for several generations of Google’s TPUs, and analyzes the results within the scope of AI-based EDA tools by Synopsys, Cadence, and academia. Finally, the limitations are evaluated, and within the next decade, an Agentic EDA, where fully autonomous Agents optimize the entire physical design flow, is proposed. Based on the scope of today’s technology, we state the true value of AlphaChip design is the ability to perform a task that, to this point, could solely be done with the aid of a computer, in a manner that can partially be taught.

DOI: http://doi.org/10.5281/zenodo.20731145

Structural Design And Optimization Of A Basement + Five-Storey Reinforced Concrete Commercial Building Under Variable Soil Conditions And Sustainability Constraints

Authors: Farah Naz, Zaheer Ahmed, Ali Hamza, Hasnain Ahmed, Saddam Hussain, Zeeshan Ali

Abstract: The structural performance of reinforced concrete buildings constructed on heterogeneous soil profiles is strongly influenced by soil–structure interaction, foundation selection, and load distribution mechanisms. This study presents the analysis and design of a basement + five-storey reinforced concrete commercial building founded on a non-uniform soil system consisting of medium-dense sand and soft clay under a high groundwater table. A three-dimensional finite element model was developed to evaluate structural response under gravity loading conditions in accordance with ACI 318-19 Building Code Requirements for Structural Concrete provisions. The results indicate that soil stiffness variation leads to differential settlement and redistribution of internal forces, particularly in beams and columns. A raft foundation system was adopted to improve load transfer and minimize settlement gradients, resulting in a maximum settlement of 21.3 mm and differential settlement of 6.4 mm. Optimization of structural elements reduced concrete and steel consumption by 10.5% and 13.1%, respectively, without compromising strength and serviceability requirements. The study demonstrates that integrating soil conditions with structural analysis and optimization enhances both performance and material efficiency in reinforced concrete buildings.

DOI: http://doi.org/10.5281/zenodo.20736250

IoT-Enabled Smart Home Security System Using ESP32 Nano With Real-Time Intrusion Alerts Via Telegram Bot And Automated Locking Mechanism

Authors: Soham Milind Lonkar, Arjun Vishwas Mane, Shivam Dhanaji Mane, Aditya Namdev Mali, Samidha Appasaheb Mandage, Swara Mamidwar

Abstract: Home security has grown to be a boosting issue in the current society and as such, there has been a necessity of affordable and effective security equipment to protect residential areas. In this paper, a smart home security system connected to the IoT is introduced and includes a laser module, ESP32 Nano, LDR sensor, buzzer, Telegram bot, and an electronic lock to identify intrusion and add more security. The system works based on an on-going scanning of a laser beam with the aid of an LDR sensor. Following the breakage of the laser path by an intruder, the LDR determines the disturbance and commands the ESP32 Nano to switch on the buzzer alarm, send an immediate notification through Telegram, and unlock the door automatically in order to inhibit illegal access. The application of IoT communication to the Telegram bot will guarantee that a homeowner can be alerted to any possible security breaches in real-time and respond adequately. The application of ESP32 Nano, which is a low-power microcontroller that has inbuilt Wi-Fi features, will make data processing efficient and also enable an easy remote communication with minimal power consumption. The design of the proposed system is aimed at being economically friendly to purchase and easy to install without requiring difficult wire connections or expensive hardware materials. It has a better automation, ability to have remote monitoring and ease of response as compared to traditional security systems. There is also the option of the very automatic locking that provides more layers of safety in reference to the intrusion security. The results indicate the system to be practical, reliable and scalable therefore making it a solution to consider in contemporary smart home setup. Additional visualization could be incorporated in the future such as camera support, AI support in motion mechanism, and cloud support in data storage to add more security and monitoring features.

DOI: http://doi.org/10.5281/zenodo.20736449

A Cloud-Based Hybrid Fake Product Detection Using MobileNet And Metadata Verification

Authors: P Kalyanakumar, Hareshvar S P, Rohan J

Abstract: The rapid growth of counterfeit goods, especially in online marketplaces, presents serious risks to consumer trust, brand reputation and the global economy. Manual or traditional verification methods are often slow, not scalable and open to manipulation. This paper introduces ProofKart, a cloud- hosted hybrid fake product detection platform that integrates visual classification using MobileNet and metadata verification techniques such as QR scanning and MRP validation. By combining deep learning and metadata-driven analysis, the system ensures robust, real-time counterfeit detection. The lightweight architecture is optimized for mobile compatibility and mass deployment across various user groups, including retailers, logistics providers, and end- users.

DOI: http://doi.org/10.5281/zenodo.20736629

MedVersa: A Review Of NLP And GAN Frameworks For Patient–Provider Communication

Authors: Madhuri Zawar, Sanmati Kumar Jain, Pradnya Vikhar

Abstract: Telemedicine is now an essential care delivery mode, but patients (especially in rural and resource-restricted environments) continue to experience challenges associated with access, symptom interpretation, and communication with providers. The development of Artificial Intelligence (AI), Natural Language Processing (NLP) and Generative Adversarial Networks (GANs) presents the opportunities to address these shortcomings. This review summarizes the latest papers on AI-enables telemedicine with the focus on: (i) telehealth applications augmented by AI analytics, (ii) NLP models of clinical records and chatbots, and (iii) GAN-based models of synthetic medical data. Recent literature was located by searching large digital libraries on the keyword’s telemedicine, NLP, GANs, and patient-provider communication and filtered on remote-care applications. The reviewed articles present a positive change in predicting risks, remote monitoring, imaging, and dialogue support, whereas patient-centered communication outcomes (e.g., comprehension, satisfaction, trust) are mentioned less frequently, and most studies use accuracy-oriented indicators, which are not statistically validated at the time. Judging by these gaps, we propose the concept of MedVersa as a conceptual framework, which includes GAN-based data generation with large language model (LLM) dialogue features to increase the dependability, fairness, and contextual sensitivity of teleconsultations and guide the implementation and evaluations in the future. This article contains a literature review and conceptual architecture, and does not indicate implemented prototype or new empirical findings.

DOI: http://doi.org/10.5281/zenodo.20741370

AI-Powered Mock Interview with Streaming Avatar

Authors: Yuvanshi Bhalawat, Yukti Baldua, Tulja Vishwakarma, Siya Purohit, Professor Pawan K. Gupta

Abstract: Most people struggle when getting ready for job talks-often less about knowing things, more about lacking chances to rehearse properly or hear useful reactions afterwards. Instead of waiting for perfect conditions, imagine trying out tough questions with someone who listens closely each time. That's where this idea steps in: a digital helper appears live, speaking naturally during mock rounds shaped just for you. Behind it sits smart software that reads your work history, picks up what matters, then builds queries tied directly to your background. As answers come through voice, the model reacts instantly – not stuck on script, shifting paths like humans do when curious. Each moment flows differently depending on how thoughts unfold, making space for pauses, changes, even surprises much like real rooms feel. Listening tools turn speech into words. These words get studied by smart programs made for spotting know-how, clear talking, steady voice, sharp ideas, how well someone does overall. Feedback comes after, showing what worked, what did not, where effort could make change. Old ways often stick to fixed questions or help from one person only now and then. This new path gives ongoing support, fits different needs, opens up whenever needed. Talking robots paired with instant scoring work together. Their job: ease stress before interviews, build better speaking habits.

DOI: http://doi.org/

Mapping The Transformation: A GIS Analysis Of Urban Sprawl For Sustainable Futures

Authors: Ishwar J. Bathe, Pavan N. Ghumare, Bhagyashree G. Satbhai

Abstract: Urban sprawl, characterized by the rapid and unplanned expansion of urban areas, poses significant challenges to sustainable development and environmental conservation. This research paper analyzes urban sprawl in Nashik City using LANDSAT 8 satellite imagery from Earth Explorer, covering the period from 2013 to 2023. In 2013, the land use was as follows: 31.81% built-up areas, 29.68% vegetation, 35.70% barren land, and 2.81% water bodies. The built-up area had grown to 38.15% by 2023. While the amount of vegetation dropped to 22%, the amount of barren land dropped slightly to 37.16%, and the amount of water bodies dropped slightly to 2.69%. This comparison shows significant changes in how land is used at that point to urban sprawl. This emphasizes the need for smart urban planning, protecting the environment, and long-term growth strategies.

DOI: http://doi.org/10.5281/zenodo.20742116

Experimental And AI-Based Multi-Objective Optimization Of An Ethanol-Hydrogen Reactivity Controlled Compression Ignition (RCCI) Engine For Ultra-Low Emissions And High Thermal Efficiency

Authors: Nand Kishore Chopara, Om Prakash Sondhiya

Abstract: Reactivity Controlled Compression Ignition (RCCI) combustion represents a highly promising low-temperature combustion strategy capable of simultaneously breaking the trade-off between nitrogen oxides (NOx) and soot emissions while maintaining high brake thermal efficiency. However, achieving ultra-low emissions and optimal performance across dynamic operating envelopes requires precise multi-variable coordination of low-reactivity and high-reactivity fuel blends. This study presents a comprehensive experimental investigation and parallel artificial intelligence (AI)-based multi-objective optimization of an ethanol-hydrogen dual-fuel RCCI engine setup. Port injection of an ethanol-water superheated vapor matrix serves as the low-reactivity fuel charge, dynamically supplemented by on-demand, exhaust-heated catalytic steam reforming of bioethanol to provide a high-velocity hydrogen-rich syngas stream. A baseline high-reactivity pilot fuel triggers the multi-stage auto-ignition process. The operational space is mapped using a three-factor, three-level Central Composite Design (CCD) framework spanning variations in reforming temperature (500°C–800°C), steam-to-ethanol molar ratios (2.0–6.0), and syngas volumetric fractions (0%–25%). To overcome classical multi-variable optimization limits, statistical Response Surface Methodology (RSM) is coupled with a back-propagation Multi-Layer Perceptron (MLP) Artificial Neural Network (ANN) trained via the Levenberg-Marquardt routine, achieving superior predictive correlation (R² > 0.994). Multi-dimensional Computational Fluid Dynamics (CFD) transit models developed in ANSYS Fluent elucidate in-cylinder turbulence, chemical species concentrations, and localized thermal profiles. The optimized AI-guided operating paradigm demonstrates a substantial 14.9% increase in maximum Brake Thermal Efficiency (BTE) under part-load frameworks, alongside dramatic drops of 65%–80% in carbon monoxide (CO) and 45%–60% in unburned hydrocarbon (UHC) profiles, realizing a practical, closed-loop technical pathway toward high-efficiency, zero-carbon scalable powertrain solutions.

DOI: http://doi.org/10.5281/zenodo.20742501

A Comprehensive Performance Evaluation Of YOLOv8 For Real-Time Urban Vehicle And Traffic Asset Recognition

Authors: Md Zahidul Islam Sany, Zhang Wubo

Abstract: Real-time object detection plays a foundational role in the deployment of Intelligent Transportation Systems (ITS) and modern smart city frameworks. To facilitate effective traffic management, automated intersection regulation, and active protection for vulnerable road users (VRUs), vision-based deep learning models must deliver high spatial localization accuracy without introducing heavy computational latencies. This research presents a comprehensive performance evaluation of the lightweight anchor-free YOLOv8 nano (YOLOv8n) architecture deployed on the CAVI-14 (Camera-based Vehicle Image) dataset. The network was trained over a rigorous 100-epoch horizon with an input resolution of 640 times 640 pixels. The experimental findings reveal high performance capabilities across critical urban mobility profiles, yielding an overall bounding box precision of 93.7%, a recall of 96.1%, a mean Average Precision mAP50 of 96.9%, and a stringent mAP50-95 of 78.4%. Notably, highly specialized categories specifically ambulance and bicycle achieved targeted localized mAP metrics of 99.5%. Detailed examination of confidence landscapes revealed that the network reached its optimal macro F1-score of 0.95 at a confidence setting of 0.358, achieving total processing throughput of ~208 Frames Per Second (FPS). This systematic evaluation validates that highly compact, parameterized single-stage deep neural networks can successfully meet the low-latency, high-precision constraints mandatory for edge-level smart city infrastructure.

DOI: http://doi.org/10.5281/zenodo.20743880

Internet of Underwater Things Infrastructure: An AI-Driven Hybrid Acoustic–Optical Framework for Real-World Underwater Communication

Authors: Assistant Professor Namrata D. Ghuse, Mayur D. Chaudhari

Abstract: The Internet of Underwater Things (IoUT) is an emerging paradigm that extends IoT technology to underwater environments for applications such as environmental monitoring, ocean exploration, and disaster prevention. However, underwater communication remains a major challenge due to high latency, limited bandwidth, and the difficulty of conducting sea trials. This review paper presents a comprehensive analysis of recent research developments in underwater acoustic communication (UAC) technologies that support IoUT infrastructure. Various studies highlight advancements in shared communication frameworks, reconfigurable hardware systems, and adaptive signal processing algorithms aimed at enhancing data transmission reliability and synchronization accuracy. Techniques such as satellite-based timing synchronization, hybrid acoustic–optical models, and learning-based optimization have demonstrated potential to improve performance under dynamic underwater conditions. Building upon the Shared Underwater Acoustic Communication Layer (SUACL) concept, this paper proposes an AI-driven hybrid acoustic–optical framework that enhances adaptability, synchronization, and energy efficiency in IoUT environments. The study identifies key research trends, evaluates comparative performances of existing methods, and discusses open challenges including high energy consumption, complex channel modeling, and security issues. Finally, it outlines future research directions focused on developing scalable, intelligent, and interoperable communication layers for real-world IoUT deployment.

DOI: https://doi.org/10.5281/zenodo.20748160

Bridging the Startup Funding Gap: PitchPortal – A Crowdfunding Platform for Entrepreneurs and Investors

Authors: Harsh Gupta, Archana Banait

Abstract: The global crowdfunding and startup investment market is experiencing explosive growth, with valuations pro-jected to surge from USD 17.7 billion in 2024 to over USD 20.46 billion by 2025. This expansion, however, masks a critical and persistent challenge: the “funding gap” faced by early-stage startups. Promising ventures, particularly those outside established tech hubs or led by underrepresented founders, consistently struggle to secure capital due to investor appre-hension, inadequate visibility, and high perceived risk profiles. While platforms like Kickstarter and Ketto have democratized philanthropic and pre-order-based support, their reward or donation-based models offer no financial stake or ownership to backers, limiting both potential returns for investors and capital-raising capacity for entrepreneurs. This paper introduces PitchPortal, a transformative crowd-funding platform engineered to bridge this gap by facilitating genuine equity-based and profit-sharing investments. PitchPortal is architected as a secure, transparent, and intelligent ecosystem that moves beyond mere connection to curated empowerment. It integrates a multi-layered verification framework, combining automated KYC/AML checks with rigorous startup due diligence to build foundational trust. At its core is an AI-powered matching engine that leverages Natural Language Processing (NLP) and collaborative filtering to intelligently align startups with suitable investors based on sector interest, risk appetite, and invest-ment history. Furthermore, to ensure transactional integrity and transparency, PitchPortal incorporates blockchain technology for smart contract-enabled, milestone-based fund disbursement, creating an immutable audit trail. By synergizing these advanced technologies within a user-centric design, PitchPortal aims to de-risk the

DOI: https://doi.org/10.5281/zenodo.20748616

Fake News Detection Using AI and ML: A Comprehensive Hybrid Ensemble Approach

Authors: Namrata D. Ghuse, Rupesh Borse

Abstract: The exponential growth of digital information has created an unprecedented challenge in combating fake news, which threatens democratic processes, public health, and social stability. Traditional manual fact-checking mechanisms are fun-damentally inadequate to address the massive scale and velocity of misinformation spread across online platforms. This research presents FakeNewsDetect, an advanced automated system that leverages synergistic Artificial Intelligence (AI) and Machine Learning (ML) techniques for robust fake news classification. Our approach implements a multi-layered analytical framework combining Natural Language Processing (NLP) for deep content analysis with a novel hybrid ensemble model for superior classifi-cation performance. The system extracts comprehensive linguistic features including sentiment patterns, stylistic markers, semantic relationships, and credibility indicators. The core innovation lies in our hybrid ensemble architecture that strategically integrates Logistic Regression, Support Vector Machines (SVM), and Long Short-Term Memory (LSTM) networks through an optimized soft voting mechanism. Extensive experimentation on diverse datasets demonstrates that our proposed model achieves state-of-the-art performance with 95.7% accuracy, 96% precision, 95.5% recall, and 95.7% F1-score, significantly outperforming individual baseline models and existing approaches.

DOI: https://doi.org/10.5281/zenodo.20748825

Beyond Text: A Comprehensive Survey of Multimodal Content Moderation Architectures in Enterprise Environments

Authors: Soham S. Jadhav, Omkar N. Gadakh, Atharv S. Gaikwad, Nisha D. Patil

Abstract: Modern enterprise workflows have evolved beyond plain text, now heavily relying on a mix of audio, video, and image-based exchanges. Consequently, the traditional paradigms of content moderation are becoming obsolete. Traditional moderation tools often fail in corporate settings because they were built for public platforms, prioritizing retroactive cleanup rather than real-time privacy and prevention, and relying on static, modality specific classifiers that fail to meet the real-time, privacy-centric, and context-aware demands of modern enterprise environments. This paper presents a comprehensive review of the state-of-the art in intelligent content moderation, analysing over 15 distinct methodologies ranging from Large Language Model (LLM)based guardrails to multimodal fusion architectures. We critically examine the transition from rigid API-based moderation to dynamic “Policy-as-Prompt” frameworks, evaluating their efficacy in handling code-mixed languages, audio-visual semantics, and organizational

DOI: https://doi.org/10.5281/zenodo.20748977

Integrating Dynamic Road Condition Sensors to Enhance Adaptability of Scenario-Specific Layers in Intelligent Transport Systems

Authors: Dhanika T. Bhangale, Assistant Professor Namrata D. Ghuse

Abstract: The use of many sensors in vehicles allows us to study how road users interact. This is important for various applications in vehicle scenarios. In this context, we introduce a new training method called Sequential Training. This method divides the Neural Network (NN) layers of the Deep Neural Network (DNN) model into two sets. One set is tailored for the user, while the other is designed to work together, focusing on the road environment. We apply deep learning in situations where vehicle users, each with unique driving behaviors and styles, interact with their surroundings. We need to create specific models for each indi- vidual vehicle user in every environment. This process requires collecting relevant data to train the machine learning models. Such data collection can be expensive and, in many cases, may even be impossible. This approach aims to integrate dynamic road condition sensors, such as weather and real-time traffic data, to improve the adaptability of scenario-specific layers.

DOI: https://doi.org/10.5281/zenodo.20749197

Machine Learning for Crime Analysis and Prediction: Modern Technique and Issues

Authors: Research Scholar Shraddha J. Pawar, Baisa L. Gunjal

Abstract: Crime forecasting has now acquired more signifi- cance in improving security and directing constabulary policies. This study will examine the current state of the art in crime pre- diction using Machine Learning techniques, specifically looking at classifiers such as linear regression, logistic regression, Neural Networks, Ensemble Methods, and Hybrid Frameworks. The paper presents the most important aspects of the approaches of analyzing the data, including data preprocessing, feature selection, evaluation of the models and algorithms, and applications. Part of the problems that are coming out in the development of this technology are touched upon briefly, such as data bias, privacy and scalability and solutions are given. The results indicate that there are certain trends as multimodal integration and spatial-temporal models, which have led to the change in the prediction accuracy.

DOI: https://doi.org/10.5281/zenodo.20749428

AI-Powered Real-Time Crisis Misinformation Debunker

Authors: Sushant Khanderao Sonawane, Ritik Prakash Patil, Shriraj Suresh Mahajan, Deepali S. Suryawanshi

Abstract: The rapid spread of misleading information across digital platforms has become a significant concern, especially during critical situations such as public health emergencies, elections, and natural disasters. Traditional fact-checking ap- proaches, which depend largely on manual verification, are often too slow to respond effectively to the speed at which information circulates online. This limitation creates a need for automated systems capable of analyzing and verifying information in real time. This research presents an AI-based framework designed to identify and verify potentially false information efficiently. The proposed system combines multiple technologies, including Large Language Models (LLMs), Retrieval-Augmented Gener- ation (RAG), and Optical Character Recognition (OCR), to process both textual and visual content. By retrieving relevant information from trusted sources and analyzing it using intel- ligent models, the system generates clear and evidence-based conclusions along with confidence scores. In addition, the framework incorporates a self-learning knowl- edge structure that improves performance over time by storing validated information and learning from previous analyses. The system is also designed to provide understandable explanations for its decisions, increasing transparency and user trust. Overall, the proposed approach offers a scalable and practical solution for real-time misinformation detection across various digital platforms.

DOI: https://doi.org/10.5281/zenodo.20749866

Snap-to-Buy: Fruit Quality Detection Using AI and ML

Authors: Samruddhi Prashant Raut, Bhumika Ramesh Sabale, Khushi Tuffailahmed Shaikh, Nikita Pradip Sabale, Assistant Professor Vaishali Khandave

Abstract: Fruit quality plays a vital role in ensuring consumer satisfaction and nutritional integrity. Manual fruit inspection is often subjective inconsistent and time consuming leading to inaccurate grading and post harvest losses. Advances in artificial intelligence (AI) machine learning (ML) and computer vision have been enable to Intelligent Systems Capable of automated quality assessment through image analysis. This paper explores the concept and architecture of SNAP TO BUY- AI based fruit quality detection and recommen- dation system inspired by research on intelligent visual inspection models. The proposed system is designed to capture fruit images via a mobile camera perform preprocessing to enhance clarity extract deep visual features using convolutional neural network (CNNs) and classify fruits based on ripeness levels ripe, unripe or over ripe Based on the analysis the system provides actionable recommendations such as buy, do not buy or store for two to three days. This study reviews rel- evant literature and presents the system architecture, dataset development strategy and modern evaluation approach. The objective is to provide a comprehensive understanding of how AL and ML can revolutionize fruit quality detection to enhance consumer decision making through intelligent automation.

DOI: https://doi.org/10.5281/zenodo.20750064

Deep Learning–Enabled Timing Optimization for Scalable and Efficient VLSI Design

Authors: Assistant Professor Atul S. Chaudhari, Aayush Mahesh Kotwal

Abstract: The continuous downscaling of semiconductor tech- nology has increased the need for reliable timing estimation and optimization in modern VLSI design. Conventional timing analysis methods depend heavily on post-routing data, which provides accurate results but demands substantial computation and slows down the overall design cycle. To overcome these limitations, recent work has moved toward incorporating deep learning models into earlier stages of the flow particularly during placement to enable faster and more predictive timing assessment. Prior studies have contributed in areas such as statistical timing analysis under process variability [1] and CNN- based frameworks for routability and power prediction, including RouteNet [4], PowerNet [2], and PROS [3]. However, these approaches operate mainly after placement or during routing, which restricts their usefulness for identifying timing issues at an early stage. The DTOC-P framework [5] marks an important shift by combining deep learning–based pre-route timing prediction with optimization capabilities of commercial EDA tools. It estimates arc delays and slew values during placement, pinpoints potential critical paths, and applies targeted refinement through iterative feedback using techniques such as buffer insertion and gate sizing. Additional features like continual learning and anomaly detection further improve its adaptability and reliability. This review paper provides an in-depth discussion of deep learning approaches applied to timing prediction and optimiza- tion, examining their methodologies, benefits, and challenges in comparison with DTOC-P. The analysis underscores how early-stage prediction and intelligent automation can shorten the timing-closure process, enhance accuracy, and reduce computa- tional effort in advanced technology nodes.

DOI: https://doi.org/10.5281/zenodo.20750273

Issues, Challenges & Role of Mathematical and Statistical Analysis in Agile Software Testing

Authors: Assistant Professor Jyotsana S. Gore, Assistant Professor Komal R. Mahekar, Assistant Professor Yogita M. Ahire

Abstract: Agile software testing emphasizes rapid iterations, continuous feedback, and frequent releases, which introduce unique issues and challenges for effective quality assurance. In this context, mathematical and statistical analysis plays a critical role in improving test planning, execution, and decision-making. One major challenge is the limited availability of stable historical data due to short sprint cycles, making accurate estimation and prediction difficult. Frequent requirement changes and evolving user stories further complicate the application of traditional statistical models. Additionally, incomplete or biased test data, time constraints, and over-reliance on intuition instead of quantitative metrics can reduce the effectiveness of testing outcomes. Despite these challenges, mathematical and statistical techniques significantly enhance Agile testing practices. Metrics such as defect density, test coverage, failure rates, and mean time to detect defects support objective evaluation of software quality. Statistical methods like trend analysis, control charts, probability models, and risk-based testing help teams prioritize test cases, identify high-risk areas, and monitor process stability. Mathematical models also aid in effort estimation, test optimization, and reliability assessment. Overall, integrating mathematical and statistical analysis into Agile software testing enables data-driven decision-making, improves test efficiency, and enhances product reliability. When appropriately adapted to Agile principles, these techniques help balance speed with quality, supporting continuous improvement and informed stakeholder confidence.

DOI: https://doi.org/10.5281/zenodo.20755037

A Unified Framework for Multi-Video Summarization and Multilingual Translation Using BART

Authors: Bhakti Waghmare, Sharon Mavelil, Dr. Jasbir Kaur, Assistant Professor Ifrah Kampoo, Assistant Professor Mansi Rajapurkar

Abstract: With the increasing classic of digital streaming services, video content is a common source for people to learn and transfer knowledge. With the rise of educational videos, podcasts and tutorials, knowledge has become easier to access; however, due to their extended length, learning via long-form video can often be time-consuming and inefficient. Therefore, this research addresses the redundancy and inefficiency of long-form videos by creating an automated system to generate short, concise summaries from video transcripts, thus reducing both length and associated learning time. The proposed system uses Automatic Speech Recognition (ASR) to convert audio to text from a video source and employs the BART model along with a hierarchical chunking approach to generate coherent and meaningful short summaries. Addi-tionally, because transformer-based models have a maximum context length, the system also executes iterative summarization to handle long transcripts effectively. Moreover, for ease of use, the functionality for cross-platform translation is included in the system and supports both high-resource and low-resource languages. Finally, the overall framework architecture is modu-lar in nature, leveraging FastAPI for the back end, while lever-aging Flutter for the front-end mobile app ensuring scalability and user friendliness.

DOI: https://doi.org/10.5281/zenodo.20759401

FaceTrack: An Automated Face Recognition Attendance System with Real-Time Blink-Based Anti-Spoofing

Authors: Digambar Santosh Awale, Sanika Arvind Golatkar, Dr. (Mrs.) Jasbir Kaur, Mr. Suraj Kanal, Ms. Mansi Rajapurkar

Abstract: Traditional attendance management techniques, including manual attendance registers and fingerprint-based biometric systems, often suffer from several limitations such as time consumption, hygiene concerns, and the possibility of proxy attendance. Although contactless facial recognition systems provide a more convenient alternative, they are still vulnerable to spoofing attacks using printed images or digital displays. To address these challenges, this paper introduces *Face-Track*, a multi-stage deep learning–based framework that combines real-time facial recognition with blink-based live-ness detection to develop a secure and spoof-resistant at-tendance monitoring system. The proposed framework em-ploys a Multi-task Cascaded Convolutional Neural Network (MTCNN) for accurate face detection and facial landmark extraction. For feature representation, a ResNet50 model pre-trained on the VGGFace2 dataset is utilized to generate 2048-dimensional facial embeddings. These embeddings are then classified using a Support Vector Machine (SVM) with a Radial Basis Function (RBF) kernel for reliable identity recognition. To prevent spoofing attempts, the system integrates a lightweight Eye Aspect Ratio (EAR)–based liveness detection approach. This method monitors eye landmark movements to identify natural blinking patterns, thereby confirming the pres-ence of a real user. The entire system is implemented through a Streamlit-based web interface, while attendance records are automatically maintained using an SQLite database.

DOI: https://doi.org/10.5281/zenodo.20759901

Time-Varying Interactions Between Macroeconomic Factors And Stock Market Volatility In Developing Economies: Empirical Evidence From Wavelet Analysis

Authors: Mrs. R. Santhiya, Dr. P. Ashok Kumar

Abstract: This study investigates the time-varying interactions between macroeconomic factors and stock market volatility in two major developing economies, India and China, using annual data spanning the period 1991-2024. The increasing integration of financial markets with macroeconomic fundamentals has intensified the need to understand the economic variables influence stock market dynamics across different time horizons by using Wavelet Coherency transformation (WCT) approaches is used to examine the co-movement, lead-lag relationships, and transmission mechanisms between stock market volatility and selected macroeconomic indicators, including GDP, inflation, export, import and gross capital formation. The findings are expected to reveal substantial heterogeneity in the strength and direction of relationships across different frequencies and economic regimes. Wavelet coherence results are anticipated to indicate stronger co-movements during periods of financial and economic turbulence, including the Asian Financial Crisis, the Global Financial Crisis, the COVID-19 pandemic, and recent geopolitical uncertainties. The phase-difference analysis is expected to demonstrate that macroeconomic variables exhibit varying leadership roles across time scales, with inflation and interest rates exerting greater influence in the short run, while economic growth and money supply dominate long-run market movements. Furthermore, the comparative analysis is likely to show that China’s stock market exhibits relatively stronger long-term coherence with macroeconomic fundamentals, whereas India’s market demonstrates more dynamic short- and medium-term interactions.

DOI: http://doi.org/10.5281/zenodo.20760756

An AI-Integrated Digital Ecosystem For Cultural Heritage Preservation, Community Engagement, And Artisan Empowerment

Authors: Dr. Pritesh Patil, Shreyas Patil, Prathvish Shetty, Ashirwad Swami

Abstract: India possesses one of the world's most diverse cultural ecosystems, comprising traditional crafts, regional languages, oral traditions, historical literature, and community-driven knowledge systems. Despite increasing digitization efforts, cultural resources remain fragmented across independent repositories, commercial marketplaces, educational platforms, and social communities. This fragmentation limits accessibility, discoverability, and long-term sustainability of cultural knowledge. This paper presents VIRASAT, a unified digital platform designed to integrate cultural preservation, knowledge dissemination, community participation, and artisan empowerment within a single ecosystem. The platform combines four interconnected modules: HeritageBazzar for artisan-centric commerce, Sangam for expert-led community discussions, Dharohar TV for multimedia cultural documentation, and Bhartiyam, an AI-powered cultural knowledge assistant based on Retrieval-Augmented Generation (RAG). The proposed architecture adopts a full-stack design utilizing React, Node.js, MongoDB, cloud-based media services, and artificial intelligence technologies to provide scalable and secure access to cultural resources. The study discusses architectural design principles, module interactions, security mechanisms, and system integration strategies that collectively contribute toward sustainable digital preservation of Indian heritage.

DOI: http://doi.org/10.5281/zenodo.20761381

Time-Dependent Comprative Analysis of Analytical and Numerical Solution for First-Order Ordinary Differential Equations

Authors: Irfan Ali, Prof: Dr. Mushtaque Hussain, Dr. Kamran Mailk, Dr. Zaheer Ahmed

Abstract: This paper presents a time-dependent comparative study between numerical and analytical solutions of first-order ordinary differential equations (ODEs), with a focus on accuracy and computational performance. Numerical methods such as Taylor Method, and the classical fourth-order Runge-Kutta method are implemented and analyzed using PYTHON. The study investigates how the accuracy of each numerical method varies with time when compared to the exact analytical solution. To demonstrate this, Newton’s Law of Cooling is used as a case study, providing a real-world application where temperature changes over time are modeled by a first-order ODE. Graphical comparisons and error analyses are carried out over different time intervals to evaluate the stability and convergence behavior of each numerical method. The results reveal that while all methods can approximate the solution, the Runge-Kutta method offers the best balance of accuracy and computational efficiency. This work highlights the importance of method selection in numerical analysis and the effectiveness of PYTHON in simulating and visualizing dynamic systems.

DOI: http://doi.org/10.5281/zenodo.20765158

A Hybrid AI Framework For Anxiety Management With Human Oversight

Authors: Ayush D. Yadav, Neelima S. Ambekar

Abstract: Anxiety disorders represent a major global mental health concern, exacerbated by limited access to professional care, social stigma, and economic constraints. Advances in artificial intelligence (AI), extensive language models (LLMs), have enabled the development of conversational agents capable of delivering scalable psychological support. However, most existing AI-based mental health systems operate in isolation from physiological, nutritional, and clinical oversight factors, limiting their clinical reliability and ethical robustness. This paper proposes a hybrid AI-driven mental health support framework that integrates an LLM-based conversational agent delivering cognitive-behavioral therapy (CBT) interventions with physiological and nutritional risk screening and structured human therapist oversight. The framework is designed to provide accessible, personalized, and ethically governed anxiety management while mitigating risks associated with fully automated systems. The paper details the system architecture, methodological design, dataset selection strategy, and evaluation metrics, positioning the proposed framework as a comprehensive and deployment-oriented model for AI-assisted anxiety management.

DOI: http://doi.org/10.5281/zenodo.20766551

Advancing FireNet-CNN For Robust , Interpretable, And Multi-Hazard Disaster Detection

Authors: Namrata D. Ghuse, Abhishek B. Bora

Abstract: Wildfires are one of the most dangerous natural disasters, causing large-scale damage to forests, wildlife, property, and human life. Early and accurate detection plays a crucial role in minimizing these losses. In this research, an extended FireNet-CNN framework is proposed for robust, interpretable, and multi-hazard disaster detection using deep learning and Explainable Artificial Intelligence (XAI). Unlike traditional review-based approaches, this work incorporates comparative experimental validation using benchmark wildfire datasets and performance metrics including accuracy, precision, recall, and F1-score. The proposed lightweight architecture is optimized for real-time deployment on edge devices, UAVs, and surveillance systems while maintaining high detection accuracy and low computational complexity. Experimental comparison with ResNet50, YOLOv8, MobileNetV2, and transformer-based models demonstrates that the extended FireNet-CNN achieves a balanced trade-off between accuracy, inference speed, and interpretability. These advancements establish FireNet-CNN as a scalable and reliable solution for real-world disaster management and intelligent wildfire monitoring systems.

DOI: http://doi.org/10.5281/zenodo.20766616

Ethical And Pedagogical Implications Of Artificial Intelligence Integration In Business Management Education: Opportunities, Challenges, And Future Directions

Authors: Dr. Ansari pulickal abdul azeez, Farooq sajjad

Abstract: The swift adoption of Artificial Intelligence (AI) in Business Management Education (BME) entails a paradoxical transformation characterized by significant opportunities for personalization and optimization, along with serious ethical and instructional threats. In this paper, we conduct a thorough and systematic evaluation of both the benefits and risks associated with integrating AI in BME, based on our multi-method research that included more than 1,500 BME students, 120 instructors, and 40 recruitment agents from 15 different institutions. Utilizing a mixed-method design, we highlight important opportunities such as a 40% reduction in grading time and a 25% increase in engagement through simulation, while identifying major threats including a 68% prevalence of student dependency among faculty and 45% of faculty reporting algorithmic discrimination issues. We recommend an innovative, two-pronged solution: the Algorithmic Ethics Compass for governance and the Augmented Intelligence Pedagogical Model for practice.

DOI: http://doi.org/10.5281/zenodo.20700957

Hybrid Human-AI Career Guidance: Opportunities, Challenges, And Future Directions

Authors: Brinceton Vaz, Khushboo Gupta, Dr. Jasbir Kaur, Mansi Rajapurkar, Ifrah Kampoo

Abstract: Artificial intelligence significantly increases the scalability and personalization in career guidance. However, the volatility in the education and job market mean these systems must stay highly relevant to a person's local economy and individual needs. This study examines AI-based career guidance systems through secondary research, synthesizing literature on recommendation systems, explainable AI, and counselling frameworks. While AI improves accessibility, skill matching, and response efficiency, its limitations in empathy and contextual judgment constrain its effectiveness for complex career decisions. Analysis indicates that a completely automated system is neither feasible nor desirable. Instead, a collaborative model where AI manages routine and data-intensive tasks while human counsellors provide interpretation, reassurance, and ethical oversight to help provide the most credible path forward. This research concludes that integrated counselling combining AI efficiency with human judgment represents the optimal direction for future career guidance systems.

DOI: http://doi.org/10.5281/zenodo.20771821

Herbal And Polyherbal Formulations In The Management Of Osteoarthritis: A Review Of Mechanistic Rationale, Standardization And Evaluation Strategies

Authors: Shruti S. Chevale, Dr. Ravi U. Kurhade, Dr. Amol D. Pagar, Amol S. Dhakpade, Nishinandan M. Shinde

Abstract: Osteoarthritis is a progressive whole-joint disorder characterized by cartilage degradation, subchondral bone remodelling, synovial inflammation, oxidative stress and functional disability. Conventional therapies provide symptomatic benefit but may have limitations during long-term use. Herbal and polyherbal formulations have attracted interest because several botanicals contain phytoconstituents with anti-inflammatory, antioxidant, analgesic and chondroprotective potential. This review summarizes the rationale for selected herbs such as Curcuma longa, Boswellia serrata, Zingiber officinale, Withania somnifera, Piper longum, Acacia nilotica, Sida cordifolia and Triphala-related formulations. It also discusses experimental models, biomarkers, histopathological endpoints and quality-by-design considerations. The review emphasizes that traditional use should be supported by authentication, standardization, marker-based quality control, toxicity assessment and controlled pharmacological or clinical evidence.

DOI: http://doi.org/10.5281/zenodo.20771964

IMHAP-Net: An Intelligent, Explainable Machine Learning Framework For Mental Health Risk Assessment And Prediction

Authors: Rohit Rana, Dr. Raj Kumar

Abstract: Background: Although the prevalence of mental health conditions like depression, anxiety, and chronic stress is on the rise among students and young adults globally, traditional clinical assessment primarily relies on self-report questionnaires and scheduled clinician contact, which can cause delays in identifying at-risk individuals. In order to support early mental health risk screening, this paper proposes IMHAP-Net, a layered, explainable machine learning framework that combines heterogeneous behavioural, academic, and self-report data with a stacked ensemble of tree-based and deep learning models. Each prediction is traceable to interpretable contributing factors. Data collection, pre-processing, feature engineering, a hybrid stacking ensemble (Random Forest, XGBoost, LightGBM, CatBoost, and a deep neural network combined via a meta-learner), an explainable-AI layer (SHAP/LIME), and a risk-assessment layer that transforms model outputs into a continuous Mental Health Risk Index (MHRI) and a four-level categorical risk label comprise the framework's six layers. To be finished following trials on the target dataset or datasets; see to Section 20 for the runnable implementation and Section 12 for the metric template.] The contributions include: (i) an open implementation path (Colab/PyTorch/Scikit-learn) that generates auditable, explainable outputs instead of a black-box label; (ii) a formally defined Mental Health Risk Index combining multiple sub-scale predictions; (iii) a stacking-ensemble formulation with explicit meta-learner equations; and (iv) a reusable six-layer architecture that separates prediction from explanation and risk scoring.

Electrical Fault Classification & Detection In Real Time By Using GAI & Deep Learning Hybrid System

Authors: Yogesh Ramesh Patni, Deepak P. Kadam, Kirti S. Kulkarni, Nilesh P. Dabe

Abstract: Real-time detection, classification, and localization of electrical faults are essential for fast protection and reliable operation of power systems. This paper presents a Diffusion-Enhanced Transformer (proposed hybrid model) that fuses a ResNet-like feature extractor and Transformer-based sequence learner with a diffusion-model generative module for data augmentation and robustness. The model is evaluated on simulated IEEE-9 bus fault waveforms and benchmarked against conventional CNN, LSTM and Transformer baselines.Experimental results demonstrate the framework’s strong real-time performance: per-fault detection accuracies of 99.1% (LG), 98.7% (LL), 98.3% (LLG) and 98.9% (LLL); overall classification metrics with precision/recall/F1 around 98.9% / 98.6% / 98.7% for the proposed model and confusion matrix showing diagonal values >0.97). The proposed hybrid achieves 98.6% overall accuracy in comparative tests while reducing fault-location error to 1.52 km (MAE), and 1.97 km (RMSE), outperforming ResNet and LSTM baselines. These findings verify that incorporating diffusion-based generative augmentation with a Transformer backbone yields improved generalization on sparse/high-noise fault data, faster inference than standard Transformers, and more accurate localization than conventional deep models, making the approach suitable for deployment in smart substations and real-time protection schemes.

DOI: http://doi.org/10.5281/zenodo.20787331

A Review On Stabilization Of Expansive Soil Using Industrial & Agricultural Wastes

Authors: Akshay R. Gurav, Pavan N. Ghumare, Swapnil. R. Joshi, Amol. B. Saner

Abstract: Expansive soils are problematic for any Construction purposes due to its Shrinkage & Swelling characteristics. Many stabilizers and Techniques have been used to stabilize expansive soils to make it appropriate for construction purposes. The selection of suitable method to address the problems caused by expansive soils is crucial, as it affects not only the soil's engineering properties but also environmental and economic factors. Theaim of this article is to review the stabilization of such expansive soils using various agricultural and industrial wastes such as SBA, Steel Slag, GGBS, FA, RHA, Saw Dust Ash, Ceramic powder, Marble Dust etc.Also, the review is focused to investigate the effect of these Stabilizers on geotechnical properties and characteristics of such soils.

DOI: http://doi.org/10.5281/zenodo.20787512

Control Strategies In Series Voltage Sag Compensator To Mitigate Inrush Current In Transformer

Authors: Prof. R. J. Nikam, Dr. D.P. Kadam, Mr. Sanjay B. Amrutkar

Abstract: The control strategies used in Series Voltage Sag Compensators (SVSCs) work to reduce the inrush current in transformers by actively managing the flux linkage in the coupling transformer. These methods commonly employ flux-oriented closed-loop control, point-on-wave switching, or fuzzy logic–based decision mechanisms. An inverter-fed series compensator injects a carefully controlled compensating voltage, ensuring smooth flux evolution during sag recovery. As a result, magnetic saturation at the instant of sag clearance is avoided, load voltage is restored seamlessly, and excessive inrush currents that could stress power electronic devices or trigger protection systems are effectively eliminated. This makes SVSCs a reliable solution for dynamic voltage compensation in modern power distribution networks.

DOI: http://doi.org/10.5281/zenodo.20787591

AI Based Intelligent Home Appliances With IoT Integration For Zero Energy Wastage And Comfort Optimization

Authors: Kunal S. Thakur, Abhilash A. Netake, Harsh M. Kharebind, Sanskruti S. Suryavanshi

Abstract: This paper proposes a hybrid hardware & software based intelligent monitoring system for conventional household appliances by integrating Internet of Things (IoT) architecture with data-driven fault detection and energy analysis. The system is designed using a dual-layer framework, where an ESP32 microcontroller performs real-time data acquisition, while a Raspberry Pi handles centralized processing and decision-making tasks. Key electrical and thermal parameters such as voltage, current, and temperature are continuously monitored to establish an adaptive baseline representing normal operating conditions. Any deviation from this baseline is analyzed using a machine learning based anomaly detection technique, enabling classification of appliance states into normal, warning, and critical categories. Unlike traditional threshold-based approaches, the proposed system incorporates trend aware analysis to detect gradual performance degradation and respond accordingly through user alerts or automatic power isolation using relay control. Additionally, real-time power and energy consumption are evaluated to assess appliance efficiency and identify abnormal usage patterns. Experimental validation on a household fan setup confirms effective detection of faults such as overheating, overcurrent, and voltage variations, while significantly reducing false alarms compared to conventional method.

DOI: http://doi.org/10.5281/zenodo.20787669

Role Of An IoT-Based Smart Energy Monitoring And Management System

Authors: Kavita Joshi, Tanay Jain, Sarvesh Khade, Jatin Patil

Abstract: The rapid growth in residential electricity consumption has intensified the need for intelligent and efficient energy monitoring and management solutions. Conventional energy meters lack real-time visibility, appliance-level insights, and automated control, leading to inefficient energy usage and increased electricity costs. This review paper presents a comprehensive analysis of IoT-based smart energy monitoring and management systems, focusing on system architectures, sensing technologies, communication protocols, data-processing techniques, and user-interface platforms. A comparative evaluation of existing systems highlights their performance, scalability, cost-effectiveness, and deployment feasibility. Key technical challenges such as data security, sensor calibration, interoperability, and communication reliability are critically discussed. Furthermore, emerging research directions including AI-driven predictive analytics, edge computing, and integration with renewable energy sources are examined. The analysis indicates that IoT-enabled energy monitoring systems significantly enhance energy transparency, reduce wastage, and support sustainable energy practices, making them a crucial component of future smart home and smart grid ecosystems.

DOI: http://doi.org/10.5281/zenodo.20787822

EM-DAT Natural Disaster Analytics: Multi-Hazard Cascading Prediction Using Machine Learning And Sequence Mining

Authors: Indrayani Subhash Shinde, Abin Mathew Mavelil, Dr. Jasbir Kaur, Ifrah Kampoo

Abstract: Between 2000 and 2025, the EM-DAT database recorded over 16,497 natural disaster events affecting billions of people and generating trillions in economic losses across 220+ countries. Traditional disaster management systems focus primarily on historical reporting and isolated hazard analysis, limiting proactive preparedness and cascading risk assessment. This paper presents the EM-DAT Natural Disaster Analytics Platform, integrating Azure SQL, Streamlit, sequence mining, Markov-chain transition analysis, and XGBoost multiclass classi-fication to analyze multi-hazard relationships and estimate next-stage disaster transitions across six analytical modules. Evaluated on 16,497 records spanning 2000–2025, the platform achieves 76.4% accuracy and a weighted F1-score of 0.74 under a 30-day cascading window, outperforming Random Forest and Naive Bayes baselines. Results are contingent on EM-DAT data quality and the chosen cascading window definition. The frame-work identifies historical multi-hazard sequence co-occurrence patterns and supports evidence-based disaster preparedness and resilience planning.

DOI: http://doi.org/10.5281/zenodo.20792498

Token Based Reconfiguration In Graph Theory: Complexity, Connectivity, And Applications

Authors: Shalini K, Dr S.P. Reshma

Abstract: The reconfiguration framework provides a unified perspective on a wide range of combinatorial and graph-theoretic problems by modeling transformations between feasible solutions under defined adjacency rules. Using the classical 15-puzzle as a motivating example, the framework extends to source problems such as independent set, dominating set, graph coloring, satisfiability, and degree-sequence realizations. Central computational challenges—including reachability, shortest transformation sequences, connectivity, and diameter—are examined, many of which are PSPACE-complete in general but admit efficient algorithms for restricted graph classes. Token-based dynamics such as sliding, jumping, and addition-removal serve as fundamental adjacency models, while extensions to puzzles, coloring, and SAT reconfiguration highlight structural and complexity landscapes. This study emphasizes the deep connections between discrete mathematics and practical applications in physics, robotics, and algorithmic game theory, offering a comprehensive foundation for future research in reconfiguration problems[1,2].

DOI: http://doi.org/10.5281/zenodo.20794444

Hardware Password Manager

Authors: Assistant Professor Dr. Niveditha, Anushka, Prakruthi, Pratibha S Patil, Thanushree

Abstract: Password security is a major concern in today's digital world. This project presents a Hardware Password Manager using the ESP32 microcontroller. The system generates secure random passwords using the ESP32's Hardware Random Number Generator (HRNG), encrypts them using the XTEA algorithm, and stores them in EEPROM memory. The encrypted passwords are transmitted through UART communication and decrypted on a computer using a Python application. The proposed system provides a secure, reliable, and low-cost solution for password management

DOI: https://doi.org/10.5281/zenodo.20795662

Disease Prediction Using Bidirectional LSTM Networks with Attention Mechanism for Symptom-Based Classification

Authors: Sonu Kumar, Omvir Singh, Ankush Kumar

Abstract: Accurate disease diagnosis from patient-reported symptoms remains a significant challenge in modern healthcare due to the complexity of symptom interpretation and the growing volume of unstructured medical text. This paper presents a hybrid deep learning model that integrates Bidirectional Long Short-Term Memory (BiLSTM) networks with an Attention Mechanism to classify diseases from free-text symptom descriptions. The BiLSTM component captures sequential and contextual dependencies in symptom narratives, while the attention layer dynamically prioritizes the most diagnostically relevant terms, improving both predictive accuracy and interpretability. The model was trained on the Symptom2Disease dataset following advanced text pre-processing, including lemmatization, stop-word removal, and TF-IDF/tokenized sequence representation, with class-weight balancing to address label imbalance. Evaluated using stratified 5-fold cross-validation, the proposed model achieved a mean accuracy of 90.51% (σ = 1.81%), and a final accuracy of 91.83% after training on the complete dataset. The model outperformed benchmark methods including k-Nearest Neighbours (81.3%), Random Forest (85.5%), and a standard Recurrent Neural Network (88.1%). These results demonstrate that combining sequential modelling with attention-driven feature prioritization yields a robust, interpretable, and clinically promising tool for automated disease prediction, with potential applications in telemedicine and clinical decision support.

DOI: http://doi.org/

Synthetic Medical Records Using Generative Adversarial Networks (GANs)

Authors: Pooja Dhankade, Dr. Pravin Kumar Malviya

Abstract: The quick data driven technology adoption of the health care sector has extinguished the notion that we require large volumes and superiorities of medical data to drive the machine learning, predictive analytics and clinical decision support systems. Simultaneously access to actual patient data is much more of an issue which we possess because of privacy laws, ethical concerns and organizational concerns. Researchers and practitioners are therefore the ones that are struggling immensely in development and validation of health care models that utilize real world data. Synthetically generated medical data has put forth as a workable and private solution to these issues. Here we put forth a model which we have named Generative Adversarial Networks (GANs) for the generation of artificial medical records out of structured and systematic health care data. We have designed the framework around practical aspects of the system, pre-processing of the data set, GAN architecture implementation, training protocols, and also the quantitative assessment. We used real medical follow up data in CSV files to train the GAN model which in turn generated synthetic data to very much like that of the real data set in terms of its statistical properties and the relationship between variables. We reported very good correlation between real and synthetic data distributions using primary clinical variables which in turn proved the put forth method’s performance. Also in general the system we present is a scalable solution to improve health care analytics data collection and at the same time it protects patient privacy which in turn is very beneficial for health care research and machine learning applications.

DOI: http://doi.org/10.5281/zenodo.20796896

Investigation Of Inter-Storey Drift And Deflection Control In Plan & Vertical Irregular Buildings Using Shear Walls, Dampers, & Base Isolators

Authors: Amool, Shivakumar kalgi, Sharankumar S. Gobbr, Brijbhushan S, Maneeth P D

Abstract: Irregular buildings are highly vulnerable to seismic and wind-induced lateral forces due to unsymmetrical stiffness and mass distribution, often resulting in excessive displacements, drifts, and uneven axial load transfer. This study evaluates the effectiveness of advanced structural control systems—shear walls, viscous dampers, and combined damper–isolator systems—in improving the lateral performance of irregular reinforced concrete buildings. A G+10 irregular RC frame was modeled and analyzed in ETABS under IS 1893:2016 provisions. The performance parameters considered include story displacement, drift, axial loads, and base shear. Results show that the shear wall configuration, though reducing axial forces, produces the maximum displacements and base shear, rendering it unsuitable for irregular layouts. The damper-only system achieved the best performance, reducing displacements by up to 41% and keeping drift values well within IS code limits. The damper–isolator system showed a balanced response but introduced localized axial load spikes, requiring careful detailing. Overall, viscous dampers perform as the most effective and practical retrofit strategy for irregular buildings, ensuring enhanced safety and serviceability without significantly increasing seismic demand.

DOI: http://doi.org/10.5281/zenodo.20801224

The Role Of Artificial Intelligence In Shaping Consumer Trust And Brand Perception In Digital Marketing

Authors: Mr. I. Prakash Raj, Dr. K. Majini Jes Bella, Dr. S. Vanitha

Abstract: Artificial Intelligence (AI) has emerged as a transformative technology in digital marketing, enabling organizations to deliver personalized, efficient, and data-driven customer experiences. The increasing use of AI-powered tools such as chatbots, recommendation systems, predictive analytics, and automated customer support has significantly influenced the way consumers interact with brands in the digital environment. As consumer trust and brand perception play a crucial role in determining business success, understanding the impact of AI on these factors has become increasingly important. This study examines the role of Artificial Intelligence in shaping consumer trust and brand perception in digital marketing through a comprehensive review of existing literature and secondary data sources. The research adopts a descriptive and conceptual approach by analyzing scholarly articles, industry reports, and recent studies related to AI applications in marketing. The review identifies key dimensions of AI-driven marketing, including personalization, responsiveness, transparency, and customer engagement, that contribute to improved consumer experiences. The findings indicate that AI technologies positively influence consumer trust by providing relevant recommendations, timely responses, and customized interactions. Enhanced trust subsequently contributes to favorable brand perception, increased customer satisfaction, and stronger brand loyalty. However, concerns regarding data privacy, security, and ethical use of AI continue to influence consumer attitudes toward AI-enabled marketing practices. The study highlights that organizations that adopt transparent and responsible AI strategies are more likely to gain consumer confidence and establish a positive brand image. The study concludes that Artificial Intelligence serves as a strategic tool for strengthening consumer trust and enhancing brand perception in digital marketing. Effective and ethical implementation of AI can help organizations build sustainable customer relationships and achieve long-term competitive advantage in the evolving digital marketplace.

DOI: http://doi.org/10.5281/zenodo.20801352

College Placements 2023 Analysis

Authors: Muhammed Yasir, D. Arnav, Nikhil Adelli, Dr. Diana Moses

Abstract: This project focuses on the analysis of a college placement dataset that captures detailed information about campus recruitment activities across multiple institutions, companies, and regions. The dataset includes key attributes such as college names, recruiters, salary packages, job domains, locations, and time-related information, allowing a comprehensive understanding of placement trends and patterns. The primary objective of this study is to explore how placement opportunities vary based on different factors such as company type, region, and institutional performance, and to derive meaningful insights from the data.The dataset was preprocessed and analyzed to ensure consistency and accuracy, followed by visualization using Tableau. Various visualizations such as bar charts, maps, donut charts, funnel charts, and waterfall charts were used to represent placement distribution, salary trends, and company participation. These visual tools helped in simplifying complex data and identifying patterns such as the variation in salary packages and the dominance of certain companies in recruitment. The analysis shows that salary packages range approximately from 4 LPA to 24 LPA, with higher packages generally offered by product-based and multinational companies. The findings also highlight that placement outcomes are influenced by regional and institutional factors, with certain colleges and locations consistently performing better. The presence of multiple job domains and work modes such as onsite, remote, and hybrid further reflects the diversity and evolving nature of the job market. Overall, this project demonstrates the effectiveness of data visualization and analysis in understanding campus placement trends and supports data-driven decision-making for students, institutions, and recruiters. In addition, this project demonstrates how structured datasets combined with visualization techniques can support better decision-making and strategic planning. Institutions can use such analysis to improve placement performance, students can gain clarity on career opportunities and salary expectations, and recruiters can better understand hiring trends across regions. Overall, the study emphasizes the value of data-driven approaches in enhancing the efficiency and effectiveness of the campus recruitment process.

Real-Time Delivery Optimization Using IoT And Deep Learning For Smart Logistics Management

Authors: Dr. Pankaj Malik, Apaar Mishra, Mohit Goyal, Mohit Bajpai, Mohid Sheikh

Abstract: The rapid expansion of e-commerce and on-demand delivery services has created a need for intelligent logistics systems capable of adapting to dynamic environments. This research presents a Real-Time Delivery Optimization framework that integrates Internet of Things (IoT) technology with Deep Learning techniques to improve delivery efficiency and operational performance. IoT devices, including GPS trackers, RFID tags, vehicle sensors, and weather monitoring systems, continuously collect real-time data related to vehicle location, traffic conditions, fuel consumption, and delivery status. The collected data are processed using Long Short-Term Memory (LSTM) networks for accurate delivery time prediction and Deep Reinforcement Learning (DRL) algorithms for dynamic route optimization. The proposed system enables real-time decision-making by continuously updating delivery routes based on changing traffic and environmental conditions. Experimental evaluation was conducted using logistics datasets containing GPS traces, traffic information, weather records, and delivery history. Results demonstrated that the proposed framework achieved 95.2% ETA prediction accuracy, reduced average delivery time by 32%, improved route efficiency by 21%, and decreased fuel consumption by 18% compared to conventional routing methods. Furthermore, customer satisfaction increased by approximately 17% due to improved delivery reliability and timely updates. These findings indicate that the integration of IoT and Deep Learning provides an effective solution for intelligent logistics management and real-time delivery optimization in modern transportation networks.

DOI: http://doi.org/10.5281/zenodo.20802364

Enhancing Agricultural Extension for Food Security: The Role of Development Communication in Kogi State, Nigeria

Authors: Obaje Friday Bernard

Abstract: Agricultural extension is a cornerstone of food security in Nigeria, yet rural communities often face barriers to accessing timely, relevant, and actionable agricultural information. This study investigates the role of development communication in enhancing agricultural extension delivery for food security in Kogi State, Nigeria. Drawing on the Participatory Development Communication (PDC) model and Diffusion of Innovations (DOI) theory, the research examines how tailored communication strategies can improve the effectiveness of extension services in disseminating agricultural innovations, strengthening farmer capacity, and fostering community participation. Using mixed methods, the study integrates recent Kogi State agricultural statistics—showing that the state’s extension agent-to-farmer ratio remains critically low at 1:3,000, with only 37% of farmers having regular contact with extension officers. Findings highlight significant gaps in extension coverage, digital inclusion, and participatory planning, while also identifying best practices from ongoing state-led and donor-supported programmes. The paper concludes that integrating development communication principles into extension delivery can significantly improve adoption rates of agricultural innovations, thereby boosting productivity, income, and food security outcomes.

DOI: http://doi.org/10.5281/zenodo.20807574

A Single Switch Isolated DC-DC Boost Converter For Extracting Maximum Power From The Solar PV System

Authors: Aarthi Bala K R, Dheebiga R, Nithya K, Vidhyasri G

Abstract: Solar PV systems face a key challenge: their output power keeps changing with sunlight and temperature. To deal with this, we developed a single-switch isolated DC-DC boost converter paired with an MPPT algorithm to pull maximum power from a PV panel under varying conditions. The converter uses a flyback-based topology with just one MOSFET switch, keeping the design simple and affordable. An Arduino Nano runs the Perturb and Observe algorithm, constantly checking PV voltage and current and adjusting the PWM duty cycle to stay at the maximum power point. We tested the system in MATLAB/Simulink and converted a 25V PV input to 200V output, achieving a voltage gain of 8 at 50 kHz switching frequency. Converter efficiency came out to 94.8% with MPPT tracking efficiency of 99.2%. The result is a low-cost, practical solution for standalone DC power applications using solar energy.

DocuMate- An Intelligent Framework For Automated, Context-Aware Documentation In Version-Controlled Software Development

Authors: Zaid Mohammad Rafique Patel, Amir Aneesh Khan, Dr. Jasbir Kaur, Ifrah Kampoo, Mansi Rajapurkar

Abstract: The rapid evolution of agile software devel-opment practices has generated increasingly complex codebases, creating a persistent gap between active source code and its corresponding technical documen-tation. This research investigates the application of generative artificial intelligence directly within ver-sion control workflows, examining methodologies for automating the extraction, generation, and continu-ous maintenance of context-aware project documenta-tion. Through the development of DocuMate—an event-driven framework integrating Large Language Mod-els (LLMs), structural project analysis, and native Git hooks—this study evaluates the effectiveness of auto-mated configuration management for critical project files, including READMEs, Dockerfiles, and environ-ment templates. Our implementation demonstrates that combining local codebase heuristics with dynamically prompted LLM generation achieves highly accurate, repository-specific documentation, significantly reduc-ing manual developer overhead compared to traditional static templating. Furthermore, the research addresses critical challenges including "documentation drift," context preservation, and deployment safety through a human-in-the-loop "pending-to-approved" state mecha-nism. The implications of this work extend to technical debt reduction, open-source maintainability, automated infrastructure configuration, and streamlined devel-oper onboarding processes.

DOI: http://doi.org/10.5281/zenodo.20808557

Behavioral Bias Modeling in Retail Investors Through Explainable Machine Learning Techniques

Authors: Research Scholar N. Rajarajeswari, Assistant Professor A. Subha

Abstract: Retail investors often display behaviourally induced biases such as overconfidence, loss aversion, and herding tendencies, which result in poor financial choices. Existing frameworks lack the capability of modelling the complex and contextual nature of these biases. In this paper, we suggest a framework for modelling and interpreting behaviourally induced biases using XML through trading logs of 5,000 retail investors. In this work, we apply Random Forest (RF), XGBoost, and a customized hybrid attention-based neural network, along with SHapley Additive exPlanations (SHAP) and Local Interpretable Model-agnostic Explanations (LIME) methods to achieve interpretability. Our approach obtains an accuracy rate of 89.2% in bias prediction compared to logistic regression (74.5%). Our results show that the highest contribution to bias comes from loss aversion and recency. Comparison studies prove the superiority of our proposed framework over existing frameworks in terms of transparency.

DOI: https://doi.org/10.5281/zenodo.20810474

Chemically Engineered Nanocomposite Membranes for Sustainable Water Purification Technologies

Authors: Dr. S. Luna Eunice, Sivakumar Neethivel, Pranav Muthuvel.B, S. Harinath

Abstract: Water pollution is becoming increasingly common and requires efficient and energy-saving methods for water purification. The current work proposes a membrane filtration system that employs chemically fabricated nanocomposite membranes composed of functionalized graphene oxide (f-GO) and silver nanoparticles (AgNPs) in the polyvinylidene fluoride (PVDF) matrix. The chemical modification (glutaraldehyde cross-linking and functionalization with carboxylic acid groups) provides additional hydrophilic, antifouling, and antimicrobial properties to the membrane. The developed nanomembrane is characterized by 325 L/m²·h pure water permeability, 99.2% bovine serum albumin (BSA) rejection rate, and 98.7% antibacterial effectiveness against E. coli. The comparative analysis of the membranes' performances against pristine PVDF and non-functionalized GO-based membranes indicates that there was a 40% decrease in the flux drop rate and a 50% increase in tensile strength.

DOI: https://doi.org/10.5281/zenodo.20810801

Blockchain-Integrated Transparent Supply Chain Framework for Counterfeit Prevention

Authors: Dr.A.Parameshwari, Assistant Professor Dr. M.Soorya Praba

Abstract: The emergence of fake products in various global supply chains causes significant economic, brand reputation, and safety implications to occur. This paper offers an innovative blockchain-enabled transparency solution for tracking the flow of goods and their authenticity using Hyperledger Fabric and InterPlanetary File System (IPFS). As opposed to the use of a centralized database approach, the architecture of the proposed system is based on smart contracts to verify the genuineness of all operations during the logistics journey, including sourcing the raw material up to reaching the point of sale. The method involves the application of hashes for products by combining the use of QR-NFT hybrid and consensus protocols. Quantitative analysis of the tested system resulted in a detection rate of 99.8%, reducing verification latency by 78% compared to other legacy solutions, and enabling instant tracing for 10,000 nodes within a second.

DOI: https://doi.org/10.5281/zenodo.20810992

Spam Detection From Organizational Emails Using Machine Learning

Authors: Sumit Ghimire, Bhoj Raj Ghimire

Abstract: Email communication plays a critical role in modern organizations. The increasing volume of spam, phishing, and malicious emails poses significant security and productivity challenges. This study investigates the effectiveness of machine learning techniques for organizational email spam detection using machine learning-based classification approaches. Enron Email Dataset and a labeled spam collection dataset containing 5,572 messages categorized as ham and spam. The data were preprocessed through text normalization, tokenization, stop-word removal, and feature extraction using sequence encoding and padding techniques. Four deep learning architectures Recurrent Neural Network (RNN), Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and Temporal Convolutional Network (TCN) were implemented and optimized. Model performance was evaluated using accuracy, precision, recall, F1-score, ROC-AUC, training time, CPU utilization, and memory consumption. Experimental results showed that all models achieved high classification performance, with accuracy exceeding 97%. The findings indicate that deep learning approaches provide effective and practical solutions for organizational email spam detection by balancing classification accuracy and computational efficiency.

DOI: http://doi.org/10.5281/zenodo.20811672

Development And Evaluation Of Herbal Oil-Loaded Hair Powder For Anti-Dandruff And Scalp Conditioning

Authors: Akanksha Santosh Sonawane, Profiessor P. S. Ghadge, Dr. Hemant Vinayak Kamble

Abstract: Dandruff is a common scalp disorder associated with fungal infection, irritation, itching, and flaking of the scalp. Conventional hair oils used for conditioning often increase scalp greasiness and may worsen dandruff conditions by providing a lipid-rich environment for Malassezia fungi. Therefore, the development of an herbal oil-loaded hair powder using a solid-state adsorption technique offers a non-greasy, effective alternative. In this study, coconut oil and clove oil were successfully converted into a free-flowing powder utilizing natural adsorbents such as bentonite, magnesium carbonate, talc, and corn starch. The formulation was further enriched with herbal powders including Azadirachta indica (Neem), Hibiscus rosa-sinensis, Emblica officinalis (Amla), and Trigonella fioenum-graecum (Fenugreek) for their synergistic antifungal, antioxidant, and conditioning properties. The prepared formulations were evaluated for various parameters including organoleptic properties, pH, angle of repose, bulk density, stability, and antifungal activity. The prepared hair powder showed excellent flow properties, acceptable skin-compatible pH, satisfactory stability, and effective anti-dandruff activity. The study concludes that the herbal oil-loaded powder is a promising, safe, and cost-effective topical preparation for the management of dandruff and scalp conditioning.

Explainable Credit Card Fraud Detection Using LightGBM And SHAP-Guided Feature Selection: A Review

Authors: Vijay Saini, Jitender Kumar

Abstract: Digital payment ecosystems have expanded both the volume and complexity of transaction data, widening the opportunity for fraud while raising customer, operational, and regulatory expectations that automated decisions remain explainable. This review consolidates research on credit card fraud detection spanning rule-based systems, classical statistical learning, ensemble and gradient-boosted methods, deep learning architectures, and explainable artificial intelligence (XAI), and synthesises these strands into FraudDetectNet, a layered pipeline in which data preprocessing, feature reduction, classification, explanation generation, and monitoring are treated as interdependent rather than separable functions. Particular attention is given to Shapley Additive Explanations (SHAP), used here not only as a post-hoc diagnostic but as an upstream feature-selection mechanism that compresses high-dimensional transaction data while preserving discriminative power. The review also examines evaluation practice for severely imbalanced, temporally ordered fraud data, arguing for precision-recall and cost-sensitive measures over plain accuracy. Outcomes reported in the primary reference study underlying this review — feature-space reduction from 380 to 120 variables, 97.6% accuracy, 99.0% fraud-class recall, and training-time reduction from 185.3 to 54.1 seconds — are presented as evidence of feasibility rather than generalised guarantees, and extensions toward graph-based, federated, and continual learning are outlined.

DOI: http://doi.org/10.5281/zenodo.20813647

AgriCare A Hybrid Cloud-Edge AI Framework For Context-Aware Crop Recommendation In Flutter

Authors: Disha Mudgal, Amey Bhogle, Dr. Jasbir Kaur, Ifrah Kampoo

Abstract: Smallholder farmers often make crop decisions with incomplete and delayed information about soil, season, and weather. AgriCare addresses this gap through a Flutter-based mobile application that combines cloud reasoning with on-device machine learning. The proposed system first attempts to generate structured crop recommendations using a Groq-hosted Llama 3.3 70B model, enriched with live weather and reverse-geocoded location context. If the cloud path fails because of poor connectivity, an API error, or a timeout, the application automatically falls back to a TensorFlow Lite classifier stored on the device. This hybrid design keeps the application useful in rural conditions where mobile data may be inconsistent, while still offering richer explanations whenever cloud inference is available. Prototype evaluation shows that the local TFLite path returns predictions in under 100 ms, while the cloud path produces more detailed ranked crop cards within a few seconds on stable connectivity. The paper presents the architecture, data flow, mathematical preprocessing, evaluation results, limitations, and future improvements of AgriCare as a practical agricultural advisory system.

DOI: http://doi.org/10.5281/zenodo.20826060

Information Bottleneck Theory In Multimodal AI: Principles, Architectures, And Emerging Research Directions

Authors: Akanksha Aher, Swaraj Rasam, Dr. Jasbir Kaur, Ifrah Kampoo

Abstract: Artificial intelligence today faces a fundamental challenge: modern AI systems are trained on enormous quantities of data, yet most of that data is irrelevant to the task at hand. The Information Bottleneck (IB) principle offers a powerful theoretical answer — it teaches AI to retain only what is genuinely useful for a task while discarding everything else. This survey examines how IB applies to multimodal foundation models that process images and text simultaneously, such as CLIP, BLIP-2, LLaVA, and Flamingo. We explain the theory in accessible terms, review how these models apply IB in practice, and discuss real-world benefits, applications, and open challenges. Our goal is to bridge information-theoretic foundations and engineering practice for the MCA research community.

DOI: http://doi.org/10.5281/zenodo.20826380

Teacher Learning Based IOT Wireless Network Optimization Models And Techniques

Authors: Sidra Khan, Sujeet Gautam

Abstract: Wireless Sensor Networks (WSNs) and Internet of Things (IoT) systems play a crucial role in modern smart applications such as smart cities, healthcare, industrial automation, and precision agriculture. Despite rapid advances in communication, cloud, and edge computing technologies, energy efficiency remains a fundamental challenge due to the limited power resources of sensor nodes. This work has proposed a IOT network optimization model that cluster nodes into cluster and further transfer packets from sensing ITO device to the base station. Clustering of model was done by teacher learning model as without prior knowledge algorithm cluster nodes in dynamic situations of WSN. Use of linear routing algorithm energy utilization was highly dropped. Experiment was done on various virtual environment of nodes and region. Result shows that Teacher Learning based Optimize Cluster Center (TLOCC) algorithm work has improves the life span of the ITO network.

A Real-Time Sentiment Analysis Framework For Flipkart Customer Feedback

Authors: Kadri Mohd Zafar Mohd Gause, Farooqui Raiyan Mehfooz Ahmed, Dr. Jasbir Kaur, Prof. Ifrah Kampoo, Suraj Kanal

Abstract: The growth of e-commerce platforms has produced large volumes of unstructured customer review text, motivating automated sentiment analysis. This paper presents a real-time sentiment analysis framework for Flipkart customer reviews that combines a multi-stage text preprocessing pipeline, comprising case normalization, URL and HTML removal, and Snowball stemming, with lexicon-based polarity scoring using NLTK’s VADER Sentiment Intensity Analyzer. The framework is evaluated on 2,304 customer reviews spanning 231 product listings, yielding an aggregate sentiment distribution of 60.0% positive, 11.7% negative, and 28.3% neutral, consistent with the dataset’s rating distribution. To address the absence of empirical validation in prior lexicon-based studies, the framework’s per-review classifications are benchmarked against rating-derived ground-truth labels using accuracy, precision, recall, and F1-score, and compared against the TextBlob lexicon-based library. Visual analytics, including donut charts and word clouds, support exploratory interpretation. The framework provides a scalable, validated, and reproducible baseline for real-time opinion mining on e-commerce review data.

DOI: http://doi.org/10.5281/zenodo.20827497

Strategic Retention Models for Reducing Workforce Attrition in Digital Enterprises

Authors: Dr.A.Parameshwari, Assistant Professor Dr.B.AGILA

Abstract: Attrition in the workforce is a critical strategic issue for digital businesses, where replacing a worker costs 150% of the annual salary for technology-specific positions. In this research paper, a comprehensive study on strategic employee retention models based on predictive analytics, customized employee experience, and skill-based training models is conducted. Using data collected on 3,500 employees at international IT corporations, it was found that prediction models based on the use of random forest algorithms demonstrate an accuracy of 87.3% in forecasting employees at risk, which helps identify such workers six months before resignation. The Integrated Retention Model includes people analytics for employee risks assessment, individual career paths, comprehensive onboarding up to 90 days, and continuous learning systems. According to comparative analysis, the application of AI-based retention techniques leads to a reduction of voluntary attrition rates by 31%, while retention of key talent increases by 28%.

DOI: https://doi.org/10.5281/zenodo.20828409

Cerrebro ERP: A MERN-Based Intelligent Educational Institution Management System With Real-Time Communication And AI-Powered Academic Assistance

Authors: Yousra Shaikh, Sahil Thonge, Sakshi Shinde, Sahana Allurkar

Abstract: Educational institutions need integrated digital systems to manage effectively their academic, administrative, and communication processes. Traditional management strategies lead to fragmentation of data, manual work, and lack of collaboration of all stakeholders. In this paper, Cerrebro ERP is described which represents a scalable system based on Educational Enterprise Resource Planning that has been developed with the help of MERN stack to organize institution's processes via centralized web application. The system includes all necessary functional components: student management, faculty management, attendance, examinations, fees, notifications, and analytical dashboard. In order to make education process more interesting, the application is equipped with AI-driven academic assistance component that helps students and real-time communication component which facilitates interaction of students, teachers, and administrators. The application uses role-based access control and authentication procedures to provide data integrity and controlled access to resources. The modular nature of the system allows making it scalable and manageable. It shows how web technologies, artificial intelligence, and real-time interaction can be applied together to improve management of educational institutions.

DOI: http://doi.org/10.5281/zenodo.20828446

A Mathematical Modeling Framework for Evaluating the Efficacy of Integrative Alternative Medicine in Chronic Disease Management

Authors: Assistant Professor Dilip Badrinarayan Soni, Dr. Hariom Singh Tomar

Abstract: Integrative alternative medicine (IAM) involves the use of traditional therapies together with evidence-based complementary therapies but is faced with the problem of difficulty in proving efficacy due to the complex nature of treatments involved. In this paper, a mathematical model for assessing IAM efficacy in chronic disease treatments is proposed. Three complementary techniques used to assess efficacy include: hybrid ODE model for disease, immune and therapy interaction; personalized fuzzy inference system involving multiple physiological and quality of life factors; and Bayesian hierarchical meta-analysis technique. Validation of the mathematical framework based on data of rheumatoid arthritis patients (n=4,752 from 23 studies) showed a sensitivity rate of 87.5% and a specificity of 84.2% in predicting response versus non-response patients. Comparison with the traditional statistics-based techniques showed a higher accuracy of prediction by 18-25%.

DOI: https://doi.org/10.5281/zenodo.20829379

Cost–Performance Optimization In Azure Data Pipelines: A Comparative Study Of Azure Synapse Analytics And Azure SQL Database

Authors: Amey Shinde, Ashmi Anavkar, Dr. Jasbir Kaur, Ifrah Kampoo

Abstract: Modern data engineering solutions increasingly rely on cloud-based analytical platforms to support both transactional reporting and large-scale analytical workloads. Microsoft Azure offers two prominent services for this purpose, Azure Synapse Analytics and Azure SQL Database, each suited to different cost and performance profiles. This paper presents a comparative evaluation of these two platforms within the context of a representative data engineering pipeline that ingests data through Azure Data Factory, stages it in Azure Data Lake Storage Gen2 using the Parquet format, and applies incremental loading strategies before serving curated datasets to a reporting layer. The study evaluates query performance, data ingestion speed, ETL/ELT execution efficiency, scalability behaviour, resource utilization, compute and storage cost, maintenance overhead, and suitability for data warehousing versus real-time analytics. Cost–performance optimization techniques, including dedicated and serverless SQL pools, the DTU and vCore pricing models, auto-pause configuration, partitioning, indexing, materialized views, workload management, data compression, Parquet file optimization, and PolyBase-based loading, are examined and discussed. Results, derived from representative benchmark scenarios, indicate that Azure Synapse Analytics provides superior throughput and scalability for large analytical workloads, whereas Azure SQL Database offers a more economical and operationally simpler option for moderate-scale transactional and reporting workloads. The paper concludes with practical recommendations for data engineers seeking to balance cost and performance when designing Azure-based analytical pipelines.

DOI: http://doi.org/10.5281/zenodo.20829245

MugenRider: An Adaptive Emission-Proportional Reward Mechanism For Sustainable Vehicle Mode Selection In Urban Ride-Hailing Platforms

Authors: Ishita Shetty, Desmond Ferrao, Dr. Jasbir Kaur, Ifrah Kampoo, Suraj Kanal

Abstract: Urban ride-hailing platforms constitute a significant and rapidly growing source of transportation-related carbon emissions in emerging economies, yet existing commercial ap-plications provide no trip-level carbon transparency and no reward mechanism calibrated to quantified emission reduction. This paper presents MugenRider, a carbon-aware ride-hailing system in which an adaptive Eco-Coin reward currency is awarded in direct proportion to the per-ride carbon saving, computed using fuel-type-specific emission factors sourced from the GHG Protocol and the Handbook Emission Factors for Road Transport (HBEFA) standards, benchmarked against an internal combustion engine (ICE) reference of 180 g CO2/km. The system is implemented as a cross-platform Flutter mobile application backed by Firebase Cloud Firestore, with route distances resolved via the Google Directions API and EV emissions computed using the Indian grid well-to-wheel factor of 0.708 kg CO2/kWh. A fixed monetary redemption equivalence of R5 per 100 Eco-Coins establishes a financially legible incentive instrument, distinguishing the system from prior non-redeemable gamification approaches. A multi-horizon EcoZone Dashboard delivers environmental feedback at immediate, periodic, and longitudinal temporal scales, complemented by a peer-comparison leaderboard. A mixed-methods empirical evaluation comprising (i) a structured functional validation suite of 33 test cases across seven system modules, all achieving 100% pass rates, and (ii) a user survey administered to 62 respondents via Google Forms, demonstrating system correctness, negligible emission computation deviation, and strong user preference for reward-linked sustainable ride selection. EV rides yielded a computed 41.0% carbon reduction relative to the ICE baseline. Five UI/UX design guidelines for reward-integrated sustainability applications are derived from the results.

DOI: http://doi.org/10.5281/zenodo.20829278

Bias In Machine Learning Models: Sources, Impacts, And Mitigation Strategies

Authors: Devansh Shukla, Dr Kaneez Zainab

Abstract: Machine Learning (ML) has emerged as a fundamental technology driving intelligent decision-making across numerous sectors, including healthcare, finance, education, e-commerce, transportation, and public administration. The increasing reliance on machine learning systems has improved efficiency, automation, and predictive capabilities. However, concerns regarding bias in machine learning models have gained significant attention in recent years. Bias can arise from various stages of the machine learning lifecycle, including data collection, preprocessing, feature engineering, model development, and deployment. Such biases may lead to unfair outcomes, reinforce existing social inequalities, and adversely affect individuals belonging to underrepresented groups. The consequences of biased machine learning systems extend beyond technical inaccuracies and can influence employment opportunities, credit approval decisions, medical diagnoses, and criminal justice outcomes. This study examines the major sources of bias in machine learning models, evaluates their social and operational impacts, and proposes a Bias-Aware Machine Learning Framework (BA-MLF) for identifying, measuring, and mitigating algorithmic bias. The framework integrates fairness assessment, bias detection mechanisms, fairness-aware learning techniques, and continuous monitoring strategies. The proposed approach aims to improve transparency, accountability, and fairness while maintaining model performance. The findings highlight the importance of responsible artificial intelligence development and provide practical recommendations for reducing bias in real-world machine learning applications.

Context-Aware Adaptive Fido Authentication Framework (Caaf): Beyond Binary Assertions for Phishing-Resistant Zero-Trust Systems

Authors: Kritika kumari ojha, Dr. Rakesh Kumar yadav

Abstract: The rapid adoption of cloud services, remote work environments, and distributed enterprise applications has fundamentally transformed modern cybersecurity requirements. Traditional authentication mechanisms based on passwords and static credentials have become increasingly vulnerable to phishing attacks, credential theft, replay attacks, and account compromise. Although Fast Identity Online (FIDO) authentication standards have significantly improved authentication security through public-key cryptography and phishing-resistant authentication, current implementations primarily rely on binary authentication assertions that either approve or deny access requests. Such static authentication decisions may not adequately address dynamic threat conditions present in modern Zero-Trust environments. To overcome these limitations, this study proposes a Context-Aware Adaptive FIDO Authentication Framework (CAAF) that integrates contextual intelligence into authentication and authorization processes. The proposed framework evaluates multiple contextual attributes, including device trustworthiness, user behavior, network characteristics, geolocation, and threat intelligence indicators, to generate adaptive authentication decisions. A Contextual Trust Score (CTS) model is introduced to quantify authentication risk and support continuous verification. The framework aims to enhance phishing resistance, reduce unauthorized access, and strengthen Zero-Trust security architectures while maintaining user convenience. The proposed approach contributes to the development of intelligent and resilient authentication systems capable of responding dynamically to evolving cyber threats.

Adaptive Perceptual Spectral Subtraction For Single-Channel Speech Enhancement With Musical-Noise Suppression

Authors: Kapil Dev Tyagi

Abstract: Single-channel speech enhancement remains a core problem in hands-free communication, hearing assistance, and automatic speech recognition front-ends, where only one corrupted observation is available. Classical spectral subtraction attains high segmental signal-to-noise ratio but injects annoying "musical noise", whereas decision-directed Wiener filtering smooths the spectrum at the cost of speech attenuation. This paper proposes Adaptive Perceptual Spectral Subtraction (APSS), a short-time spectral gain that unifies four ideas: a decision-directed a priori signal-to-noise ratio driving a parametric square-root Wiener gain that preserves speech, a soft speech-presence indicator derived from the a posteriori SNR, a masking-guided suppression stage that removes only inaudible residual energy, and an SNR-adaptive temporal smoothing of the gain that erases isolated spectral peaks without smearing onsets. The method is evaluated on controlled synthetic speech corrupted by white, pink and babble noise from -5 to 15 dB. On stationary noise APSS attains the highest segmental SNR (7.9 dB, versus 7.8 dB for spectral subtraction and 4.5 dB for Wiener) while reducing log-spectral distortion by roughly 9% relative to spectral subtraction, and it keeps the musical-noise kurtosis ratio about seven times lower than the Wiener baseline. The results show that APSS occupies a favourable point on the distortion-versus-musical-noise trade-off surface that neither baseline reaches.

DOI: http://doi.org/10.5281/zenodo.20838131

Machine Learning-Based Clustering And Prediction Of Students Mental Health Care Needs: Evidence From Hanoi National University Of Education

Authors: Hoang Thi Lam, Pham Minh Phuong, Pham Quoc Thang

Abstract: This study extends an undergraduate thesis dataset on mental health care needs among 300 students at Hanoi National University of Education by adding machine learning analyses. Three need dimensions, namely cognitive/informational needs, emotional support needs, and access-to-service needs, were used for K-means clustering, while demographic variables, DASS-21 scores, and contextual factors were used to train Decision Tree and Random Forest models. Results showed a relatively high overall need level (M = 3.93/5). K-means with k = 3 identified low-need (25.7%), moderate-need (44.0%), and high-need (30.3%) groups. Random Forest outperformed Decision Tree, reaching 78.9% accuracy in three-class prediction and 85.6% accuracy in detecting high-need students. Family, learning-environment, socio-cultural, and personal factors were the strongest predictors. The findings suggest that machine learning can complement traditional statistics for early screening and targeted student support.

DOI: http://doi.org/10.5281/zenodo.20838284

Influence of Machining Parameters and Optimization Techniques in CNC Milling Operations

Authors: Sangamnere Swapnil, Waghchaure Pratik, Musale Sidhant, Mogal Prashant, Sachin Kakade

Abstract: Computer Numerical Control (CNC) milling is a precision machining process used to manufacture complex and accurate components by controlling machine tools through programmed instructions. It plays a vital role in modern manufacturing industries due to its repeatability, flexibility, and high dimensional accuracy. CNC milling performance mainly depends on process parameters such as spindle speed, feed rate, depth of cut, and tool geometry. From an expert manufacturing perspective, CNC milling combined with Computer-Aided Manufacturing (CAM) enhances productivity by optimizing tool paths and machining parameters before actual production. The use of CAM software reduces trial-and-error, minimizes tool wear, and improves surface quality. This integration is especially beneficial when machining medium carbon steels like EN8, which are widely used in automotive and mechanical applications. The major problem addressed in this Research is the selection of optimal milling parameters to achieve high material removal rate with good surface finish and dimensional accuracy while reducing machining time and tool wear. CAM software is used for tool-path generation, machining simulation, and parameter optimization, as it allows virtual verification and efficient planning of CNC milling operations. The results achieved show that optimized spindle speed, feed rate, and depth of cut significantly improve surface finish, reduce machining time, and enhance overall machining efficiency for EN8 material.

DOI: https://doi.org/10.5281/zenodo.20842428

PhishShield: A Detection and Prevention System for Phishing Domains

Authors: Suyash Kangude, Prashant Aher, Aniket Mehare, Sanket Bhavari, Harshada Ahire, Abhishek Badadhe

Abstract: Phishing remains one of the most widespread and dangerous cyberattacks, where adversaries create deceptive domains and fraudulent websites to obtain confidential information from unsuspecting users. As phishing techniques evolve and become increasingly advanced, traditional security measures are often unable to recognize or block these threats, since attackers continually modify their strategies. This project introduces PhishEye a machine learning–based phishing domain detection system designed to identify, evaluate, and respond to phishing attempts. The framework follows a structured pipeline consisting of data collection, feature extraction, model training, real-time threat detection, takedown request automation, and dashboard monitoring. Datasets were gathered from trusted sources such as PhishTank and OpenPhish, complemented by additional feeds and domain permutations generated with Dnstwist. To create robust feature vectors, multiple attributes were included: lexical, domain, network, SSL/TLS, and webpage content features. Machine learning models—including Random Forest and Logistic Regression—were trained and validated using evaluation measures such as accuracy, precision, recall, and F1-score. Once deployed, the system monitors domains in real time and assigns a risk score (low, medium, or high) to estimate their legitimacy. For domains classified as high risk, PhishEye initiates an automated takedown request to the hosting provider, registrar, or relevant CERT authority. Finally, a web-based dashboard provides a comprehensive interface for administrators, showing real-time statistics and system performance for effective phishing defense.

DOI: https://doi.org/10.5281/zenodo.20842661

Design and Feasibility Analysis of a CNC-Based Friction Hardening Tool for Enhanced Wear Resistance in Stainless Steel Pressure Valve Seats

Authors: Pranav Kute, Drishika Gupta, Srushti Deshmane, Aditya Kapse, Shirsath Seema B, Dhiraj Deshmukh

Abstract: Industrial pressure valves frequently experience failure due to severe wear at the valve seat, a critical region with complex geometry. Conventional surface hardening techniques, such a induction hardening, often result in uneven hardness profiles, thermal distortion, and micro- cracking. This study proposes a novel, eco-friendly “Friction Hardening” process utilizing a custom-designed, CNC-mounted tool to achieve localized surface strengthening on stainless steel valve seats. The process utilizes frictional heat and contact pressure (approx. 15 MPa at 600 RPM) to induce severe plastic deformation and dynamic recrystallization in the surface layer. Preliminary analysis suggests this method can create a refined microstructural layer with significant hardness improvement and negligible geometric distortion, offering a sustainable alternative to traditional thermal hardening.

DOI: https://doi.org/10.5281/zenodo.20842996

TnP Vision: Unified Platform for Transparent and Skill-Centric Campus Placement Management

Authors: Thejas Raja Eladassery, Sanika Kalaskar, Aditi Jadhav, Shreya Gaikwad, Dr.V.G.Kottawar

Abstract: Many higher education institutions still have a disjointed, manual, and ineffective campus placement process, which results in low transparency, little student visibility, and less than ideal recruitment results. In order to modernise and simplify campus placement management through a centralised, skill-centric ecosystem, this paper introduces TnP Vision, a unified platform driven by artificial intelligence. Student profiles, academic verification, recruiter communication, and placement procedures are all combined into one digital platform by the system. While faculty-driven verification guarantees the legitimacy and authenticity of accomplishments, TnP Vision uses Natural Language Processing (NLP) to extract and classify skills from unstructured student data. For students, teachers, recruiters, and administrators, the platform offers role-based dashboards that allow for customised access to insights, automated portfolio creation, and structured interview feedback. Real-time analytics also facilitate institutional performance monitoring and data-driven decision-making. When compared to conventional placement systems, exper-imental evaluation shows that the suggested system greatly increases student engagement, data reliability, and candidate discovery efficiency. TnP Vision improves employability outcomes and builds recruiter trust by switching from resume-based evaluation to skill-based discovery and verified profiles. By integrating AI-driven skill identification, academic valida-tion, and integrated workflow management into a single scalable platform, the suggested solution fills a significant gap in current placement systems and offers a revolutionary approach to campus recruitment in higher education.

AutoDevOS: AI-Driven Autonomous Software Development Orchestrator using Model Context Protocol

Authors: Archana Banait, Yash Khalkar, Soham Gaikwad, Kanhaiya Sharma, Vivek Sanandiya

Abstract: AutoDevOS is an AI-driven autonomous software development orchestrator designed to revolutionize the way soft- ware projects are built and managed. It allows users to provide high-level natural language prompts, which are decomposed by a central meta-agent into specialized sub-tasks. These tasks are then handled by distinct intelligent agents (frontend, backend, DevOps, testing, and documentation). These agents collaborate seamlessly to generate, integrate, and deploy production-ready applications with minimal human intervention. The system leverages large language models (LLMs) alongside the Model Context Protocol (MCP) for dynamic context management and tool execution. Key observations from our prototype demonstrate a 40% reduction in boilerplate generation time and an 85% task completion rate for standard multi-file operations. By combining modular task handling, automated loop detection, and DevOps integration, AutoDevOS significantly improves code quality and prevents context degradation—paving the way for the future of intelligent, autonomous software engineering

DOI: https://doi.org/10.5281/zenodo.20843654

A Review of ESG Reporting Practices in Indian REITs

Authors: Professor Neha Wankhede, Dr. Namrata Deshmukh

Abstract: Environmental, Social, and Governance (ESG) considerations have become very important in the Real Estate Investment Trust (REIT) sector as investors place greater emphasis on sustainability, transparency, and long-term value creation. Despite the growing adoption of ESG reporting frameworks, significant inconsistencies persist in how REITs measure, disclose, and standardize ESG information, limiting comparability and investor confidence. This study examines the processes, frameworks, and challenges associated with ESG reporting and standardization in the REIT sector. Using a descriptive research design based on secondary data, the study reviews academic literature, regulatory guidelines, global ESG frameworks, and ESG disclosures of selected Indian REITs, including Mindspace Business Parks REIT, Brookfield India REIT, and Embassy Office Parks REIT. The findings shows that while structured governance frameworks and clear ESG policies improve reporting consistency at the firm level, the absence of uniform, REIT-specific ESG standards remains a key challenge. The study point outs the need for stronger standardization, improved regulatory guidance, and sector-specific disclosure norms to enhance transparency, reduce greenwashing risks, and support informed investor decision-making in the REIT market.

DOI: https://doi.org/10.5281/zenodo.20844005

Thermophysiological Performance Of Sheep Wool, Goat Wool And Their Hybrid Composite Fabrics

Authors: Elif Altürk, Nazlı Tatar, Zeynep Rüya Parlak

Abstract: Wool's ability to easily absorb sweat and not feel damp, keep you warm, protect against radiation, and its antibacterial properties have always been very appealing to humanity. In this study, three different fabrics were produced using two different natural fibers: goat and sheep wool. The thermal conductivity, thermal absorptivity and diffusion values of the fabrics obtained were measured. The results showed that sheep wool has the highest moisture absorption property. The measured thermal absorptivity values were 121 W·s¹ᐟ²/m²K for sheep wool, 113 W·s¹ᐟ²/m²K for goat wool, and 73 W·s¹ᐟ²/m²K for the composite structure. The maximum heat flows of sheep, goat, and sheep-goat composite fabrics were measured, and the sheep-goat composite fabric, with the lowest value of 166W/m², was the fabric that felt the warmest upon first contact. The results showed that sheep -goat composite fabric showed very good thermo-physiological performance that can be sustainable alternative compare to synthetic fabrics.

DOI: http://doi.org/10.5281/zenodo.20844460

Design and Performance Optimization of 2.4ghz Microstrip Patch Antenna for WI-FI, Wlan and Bluetooth Applications

Authors: Ishwari Gaikwad, Swapnil Kadam, Pranv Kela, Kalindi Mahajan

Abstract: The rapid expansion of the wireless communication systems has led to an immediate need for small antennas that can provide high performance and reliable operation in the 2.4GHz ISM band which is commonly used in Wi-Fi WLAN and Bluetooth systems. Microstrip patch antennas represent the optimal solution because they deliver compact design advantages and lightweight features and simplified manufacturing processes and printed circuit board compatibility. Conventional antennas however operate with restricted bandwidths which create impedance matching issues that degrade their overall transmission capacity. The study investigates the creation and improvement process of 2.4Ghz microstrip patch antenna through the application of an FR4 substrate and microstrip line feeding system. The key parameters such as patch length, patch width are optimized to enhance bandwidth and overall antenna performance. The simulation results shows that return loss has decreased to a value below -10db and VSWR reaches an acceptable level below 2 while gain operates at the needed resonant frequency. The developed design is suitable for wireless communication devices including Wi-Fi routers, Bluetooth modules and WLAN enabled systems which provide both dependable short-range communication and overall stability.

DOI: https://doi.org/10.5281/zenodo.20844492

Neural Network for Autonomous Vehicle Perception: A Review of Traffic Sign Recognition, Lane Detection, and Deployment Challenges

Authors: Research Scholar Viraj Raghunath Sonawane, Professor Balasaheb Agarkar, Associate Professor Sachin Chaudhari

Abstract: This review provides a comprehensive study of neural network-based perception pipelines for autonomous vehicles, concentrating on the accuracy-complexity trade-offs affecting Traffic Sign Recognition (TSR) and lane detecting systems. The survey outlines the progression from conventional vision algorithms to modern deep learning frameworks, emphasizing how Convolutional Neural Networks (CNNs) and analogous architectures have revolutionized robustness, feature discrimination, and real-time inference capabilities. In TSR, high-capacity architectures such as ResNet-101 and attention-augmented models demonstrate improved resilience to small-scale, occluded, and luminance-variable sign categories, while single-stage detection frameworks, exemplified by the YOLO family, offer latency-constrained inference suitable for embedded applications. Conversely, two-stage frameworks such as Faster R-CNN provide enhanced representational accuracy but entail much higher computational expenses, underscoring the intrinsic precision–efficiency trade-off characteristic of contemporary vision systems. This study broadens the inquiry into lane detection, where deep learning techniques encounter challenges related to environmental unpredictability, low-contrast features, occlusions, and sensor noise. Techniques employing structured ROI extraction significantly diminish unnecessary computation while preserving accuracy. The study combines algorithmic performance, computational constraints, and deployment viability to emphasize the importance of ongoing innovation in neural network topologies, as well as the provision of comprehensive and carefully annotated datasets for the unified approach of lane and traffic sign detection. Future research should focus on closing the gap between performance and the real time requirement of driver assistant systems, aiming for a fine mix of accuracy, efficiency, and reliability in dynamic real-world contexts.

DOI: https://doi.org/10.5281/zenodo.20844723

Computing Various Optimal Performance Trade-Offs Parameters for FBMC Based Modulation Scheme

Authors: Research Scholar Nandkishor P. Sonawane, Assistant Professor Surekha Patil

Abstract: Filter Bank Multicarrier compresses data more tightly for one reason? It bypasses the cyclic prefix step entirely. Still, that efficiency comes at a price – processors work harder even under strict constraints. based on the experiments runs the performance of develop FBMC configurations is assessed. Tracing signals from filter to delivery steps allows each trial to track BER, power spikes compared to typical baselines, and total data transported. Change things like pulse form or distance between subcarriers, performance moves in unambiguous ways. Because symbols spread out, they carry deeper detail and cut mistakes. If filtering occurs too frequently, blur begins to appear. However, the signal becomes blurry when filters are repeated too frequently. Sizing is crucial because peak power increases dramatically as symbol length increases. Choosing the proper filter repeat level is important because either extreme throws off performance. PAPR increases with sequence growth, particularly between 10,000 and 100,000 symbols. As interference decreases, speed rapidly increases; this is especially noticeable in large sets that are close to 20,000 units at maximum intensity.

DOI: https://doi.org/10.5281/zenodo.20844915

Design and Manufacturing of Pollution Monitoring System

Authors: Assistant Professor A. Y. Chaudhari, Soham Nikumbh, Sahil Pagare, Mayur Koli, Nikhil Patil

Abstract: This study shows the construction of a compact pollution monitoring system designed to measure the percentage of the gases that escape from the cylinder, like oxygen (O₂) and carbon dioxide (CO₂). These are harmful to the environment, so to reduce the percentage of pollution of air through the escape gases, a custom-built filter/catalytic converter chamber is used. In this system, for interpreting exhaust gas data, the Arduino Nano microprocessor was used to display the gas proportion in real time on the 16 x 2 LCD screen. When the level of exhaust gases exceeds the predetermined safety threshold value level, an audible siren notifies the operator, and a 10-minute countdown timer starts as a warning. If the excessive CO₂ condition persists beyond this duration, the system automatically disconnects electricity to terminate operation, thereby preventing potential risks. This prototype demonstrates useful applications in environmental safety and emission regulation by providing an affordable, real-time solution for monitoring and managing emissions from small engines or laboratory exhaust.

DOI: https://doi.org/10.5281/zenodo.20845115

Hydrogen and Nanoparticle Enhanced Waste Cooking Oil Biodiesel in Diesel Engines: A Literature Survey

Authors: Dipak K. Dond, Pushkar D. Khambekar, Kunal S. Gaikwad, Sahil R. Dhatrak

Abstract: The continuous depletion of fossil fuel reserves and the enforcement of stringent emission norms have accelerated the need for sustainable alternatives in compression ignition engines. Biodiesel derived from waste cooking oil (WCO) presents a viable solution by addressing both energy demand and waste disposal concerns. However, its inherent drawbacks—such as higher viscosity and lower calorific value compared to diesel—can negatively influence engine performance and emissions. To overcome these limitations, recent research has focused on enhancing combustion through hydrogen supplementation and nanoparticle additives. Hydrogen, due to its high diffusivity and rapid flame propagation, improves combustion efficiency, while nanoparticles act as catalysts that promote oxidation reactions. Additionally, fuel injection parameters such as pressure, timing, and quantity significantly influence combustion characteristics. This review examines previous studies on WCO biodiesel enhanced with hydrogen and nanoparticles, with emphasis on injection system optimization and its effects on engine performance, combustion behaviour, and emission characteristics.

DOI: https://doi.org/10.5281/zenodo.20845282

HerHealth AI: Intelligent Diet Planning And Yoga Posture Optimization For Holistic Well Being

Authors: Saee N. Suryawanshi, Pornima P. Sangale, Shruti K. Pawar, Shital R. Bedse

Abstract: Throughout a woman’s life, physiological and hormonal changes—such as menstruation, pregnancy, postpartum recovery, menopause, and conditions like PCOS and anaemia—significantly affect her health. Unhealthy habits and improper dietary choices often worsen these challenges. While many digital health apps exist, most offer generic recommendations and lack real-time, personalized support for women. This paper presents an AI-powered wellness system specifically designed for women, integrating personalized diet planning with yoga pose correction. The system uses machine learning for nutrition and computer vision techniques, including CNNs and MediaPipe, for pose detection and correction. It addresses limitations of existing solutions, such as lack of real-time feedback, conditionspecific guidance, and poor integration of wellness practices. By combining pose correction, personalized nutrition, and progress tracking into a single AI platform, it enhances the effectiveness and accessibility of women-centric wellness solutions.

DOI: http://doi.org/10.5281/zenodo.20854050

Context-Aware Password Management For Multi-User IoT Environments: A Systematic Review Of Secure Key Rotation And Expiry Mechanisms

Authors: Vrutti Mistry, Yassir Farooqui, Amit Barve

Abstract: The rapid proliferation of Internet of Things (IoT) devices across smart homes, healthcare systems, and industrial environments has intensified the need for robust and adaptive security mechanisms in multi-user settings. Traditional password management approaches remain widely deployed; however, they suffer from persistent vulnerabilities including weak password selection, credential reuse across services, and the absence of structured lifecycle management mechanisms. This paper presents a systematic review of existing authentication, password management, and key lifecycle strategies applicable to multi-user IoT ecosystems. The study follows a structured review methodology to analyze and synthesize contemporary research contributions in the areas of context-aware authentication, secure key rotation, password expiry mechanisms, and lightweight cryptographic implementations. A comparative evaluation of diverse security techniques—such as one-time passwords (OTPs), zero-knowledge proofs (ZKP), symmetric and public-key cryptographic schemes, and machine learning-based threat detection models—is conducted with particular attention to device resource constraints, scalability challenges, and operational efficiency. Conceptual models, analytical tables, and comparative charts are utilized to highlight trade-offs between security strength, computational overhead, and system performance. The review identifies significant research gaps in integrating dynamic key rotation and expiry mechanisms into holistic, context-aware security architectures tailored for multi-user IoT environments. Finally, the paper outlines future research directions aimed at developing scalable, resource-efficient, and adaptive password lifecycle management frameworks for next-generation IoT systems. management frameworks for next-generation IoT systems.

DOI: http://doi.org/10.5281/zenodo.20854861

Social Media Forensic: Neural Network Techniques For Cyberbullying Hate Speech Detection Under Uncertainty

Authors: Sakshi Dinesh Pandit, V. D. Dabhade

Abstract: The rapid growth in online hostile behavior, in- cluding hate speech and cyberbullying, has created significant challenges for users and platform moderators. While early detection methods relied on traditional machine learning ap- proaches like SVM, Naïve Bayes, and Random Forests using hand-crafted features [3], [4], these techniques struggled with noisy, multilingual, and imbalanced social media data. Recent advances in deep learning, particularly transformer-based models such as BERT, RoBERTa, and architectures like CNN, LSTM, and BiLSTM, have substantially improved detection capabilities by automatically learning rich textual representations [5]–[8]. However, most existing systems generate deterministic, overcon- fident predictions even in ambiguous contexts, posing risks in high-stakes moderation scenarios. This review comprehensively examines hate speech and cyberbullying detection techniques with emphasis on uncertainty-aware learning approaches, in- cluding Monte Carlo dropout, Bayesian modeling, and ensemble methods. We identify critical research gaps related to dataset bias, annotation inconsistency, limited multilingual coverage, and insufficient robustness. Our findings highlight the necessity of integrating uncertainty quantification into detection systems to enhance reliability, interpretability, and real-world applicability, guiding future development of trustworthy and ethically respon- sible content moderation frameworks

DOI: http://doi.org/10.5281/zenodo.20856177

Self-Learning Charging Networks Using Reinforcement Learning Agents For Intelligent Electric Vehicle Charging Infrastructure

Authors: Vishu, Dr Raj kumar

Abstract: The rapid proliferation of electric vehicles (EVs) is imposing unprecedented stress on existing power grid infrastructure, necessitating intelligent, adaptive, and scalable charging management systems. Conventional rule-based and heuristic scheduling approaches fail to accommodate the stochastic nature of EV arrival patterns, fluctuating renewable energy availability, and real-time grid constraints. This paper proposes the Adaptive Multi-Agent Self-Learning Charging Network (AMSLCN), a novel multi-agent reinforcement learning (MARL) framework designed to optimize EV charging operations across distributed charging station networks. AMSLCN employs a decentralized execution with centralized training (DECT) paradigm, in which each charging station hosts an independent Deep Q-Network (DQN) agent that learns optimal scheduling policies through continuous interaction with a simulated smart grid environment. The proposed framework jointly optimizes charging efficiency, energy cost minimization, user waiting-time reduction, grid stability, and renewable energy utilization. The mathematical formulation of the problem is cast as a Markov Decision Process (MDP), with carefully designed state representations, action spaces, and a composite reward function that encodes multiple operational objectives. Extensive simulation experiments, conducted using real-world EV charging datasets from the ACN-Data repository and synthetic grid load profiles, demonstrate that AMSLCN achieves a 34.7% reduction in average energy cost, a 41.2% improvement in charging efficiency, a 38.5% decrease in user waiting time, and a 29.3% increase in renewable energy utilization compared to the best-performing baseline. AMSLCN significantly outperforms rule-based scheduling, genetic algorithm-based optimization, fuzzy logic controllers, standalone Q-Learning, and single-agent DQN across all evaluation metrics, establishing it as a compelling foundation for next-generation intelligent EV infrastructure.

Optimized Design And Emerging Applications Of Arithmetic Logic Units In Modern Computing

Authors: Vijay Kumar J, Panduranga P, Chandrasekhar Reddy C

Abstract: In contemporary computer systems, the Arithmetic Logic Unit (ALU) serves as the central component of the Central Processing Unit (CPU), executing fundamental tasks such as arithmetic calculations, logical operations, and data transfer. As information technology advances, the demand for computing continues to rise, leading to increasingly stringent requirements for accuracy, speed, and energy efficiency in computing processes. The design and optimization of the Arithmetic Logic Unit (ALU) are crucial for enhancing both the performance and energy efficiency of a system. Consequently, they are central to research and technological advancements. This paper aims to summarize and investigate the application scenarios, development, and optimization potential of ALUs. It reviews various contexts, such as embedded systems, quantum computing, and high-performance processors, to illustrate how customized ALU designs can address specific requirements. In terms of optimization strategies, the study discusses advanced arithmetic circuits using 5 stages of pipeline RISC-V, next-generation logic gates, Quantum cost-optimized reversible vedic multipliers, Gate Diffusion Input (GDI) technology, reversible logic, Single Electron Transistor (SET) and Quantum Cellular Automata (QCA) technology as methods to reduce power consumption and increase processing speeds.

DOI: http://doi.org/10.5281/zenodo.20861986

Design and Development of an Educational Web Security Analysis System for Vulnerability Detection

Authors: Krishnaraj Chandrapratap Chauhan, Sabi Ahsan Sayyed, Dr. Jasbir Kaur, Assistant Professor Sandhya Thakkar

Abstract: Website security has become a critical concern due to the increasing number of cyberattacks targeting web applications. Many websites remain vulnerable because of improper security configurations such as invalid SSL certificates, missing HTTPS enforcement, insecure cookies, and absence of security-related HTTP headers. Existing security assessment platforms often require advanced technical expertise and are difficult for students and novice developers to utilize effectively. This paper presents the design and development of an Educational Web Security Analysis System capable of automatically evaluating website security configurations. The proposed system performs SSL certificate validation, HTTPS redirection verification, security header analysis, and cookie security assessment while generating security scores, visual dashboards, screenshots, and downloadable PDF reports. Developed using Python, Streamlit, Selenium, and Requests libraries, the platform aims to provide a simplified and educational approach to website security assessment. Experimental evaluation demonstrated detection accuracy exceeding 94% across all modules and an average report generation time of 2.1 seconds. The developed system achieved a System Usability Scale score of 89/100, indicating strong usability and effectiveness for cybersecurity education. Furthermore, detailed feedback from 50 student participants confirmed that the platform successfully bridges the gap between theoretical security concepts and practical vulnerability detection.

DOI: https://doi.org/10.5281/zenodo.20921571

Current Practice in Cost Estimating and Cost Control in Tendering and Bidding Process in Highway Construction

Authors: Gawai Santosh Bhaskar, Shashikant B. Dhobale

Abstract: Accurate cost estimation and effective cost control are critical challenges in highway construction tendering and bidding processes. Traditional estimation methods often rely on manual calculations and historical averages, which may lead to inaccuracies due to market fluctuations, negotiation variability, and complex project parameters. This research proposes a hybrid optimization framework integrating Regression Analysis and Genetic Algorithm (GA) to enhance prediction accuracy and optimize Total Contract Cost in highway construction projects. Initially, a regression-based predictive model is developed to estimate total contract cost using key influencing parameters such as material cost, labor cost, equipment cost, and negotiation factors. Subsequently, a Genetic Algorithm is applied to minimize the predicted cost and determine the optimal combination of decision variables under defined constraints. The model successfully generates multiple Pareto optimal solutions, providing flexible trade-off options for strategic decision-making. The MATLAB simulation results demonstrate that the proposed hybrid GA–Regression model effectively identifies the optimized total contract cost and enhances negotiation strategy performance. The findings highlight the practical importance of intelligent optimization techniques in improving bidding competitiveness, minimizing financial risk, and strengthening cost control mechanisms in highway construction projects. This research contributes a data-driven, optimization-based decision support framework suitable for modern tendering practices.

Analysis Of Finite Capacity Queueing Model With No-Passing, Reverse Balking And Reneging

Authors: Priyanka, Poonam Singh, Swadesh Singh

Abstract: In this study, we present Markovian queueing model with finite capacity, reverse balking and reneging of impatient customers. Some performance measures including queue length, waiting time, transient probabilities, some special case and optimal solution are obtained by applying numerical techniques. To verify the numerical results, computational software has been used.

DOI: http://doi.org/10.5281/zenodo.20924364

Parametric Optimization Of Abrasive Water Jet Machining (AWJM) Process For Non-Conventional Rubber-Based Materials

Authors: Sunil A Kumavat, Amol P Chaudhari

Abstract: This work presents an experimental investigation on the optimization of Abrasive Water Jet Machining (AWJM) parameters for rubber-based composite materials, which exhibit complex viscoelastic behaviour. Conventional machining methods often lead to deformation and thermal damage in such materials, making AWJM a suitable alternative due to its cold cutting nature. A Taguchi L27 orthogonal array was employed to systematically evaluate the influence of water pressure, traverse speed, and stand-off distance on surface roughness. A regression-based predictive model was developed to quantify parameter interactions. The statistical analysis confirmed that water pressure has the highest influence on surface quality, followed by traverse speed and stand-off distance. The optimal surface roughness of 2.857 µm was achieved at higher pressure levels combined with moderate-to-high traverse speed and increased stand-off distance. The results establish a parameter selection framework for efficient machining of deformable rubber materials and provide a basis for future integration of advanced predictive models

DOI: http://doi.org/10.5281/zenodo.20927704

R&D of AI Language Models for Scalable Detection & Support in IGD and BDD

Authors: Omprakash Pandey, Harshal Chavan, Anubhav Sharma, Dr. Vinayak Kottawar

Abstract: Internet Gaming Disorder (IGD) and Body Dysmorphic Disorder (BDD) represent escalating mental health conditions characterized by compulsive digital behaviors and distorted self-perception. Traditional diagnostic methodologies heavily rely on sporadic clinical assessments, limiting large-scale, real-time early detection and intervention. This paper introduces AegisMind, a novel, scalable, multimodal AI language and physiological monitoring system engineered for early risk identification and real-time psychological support. AegisMind integrates an AI Orchestrator combining an NLU engine and Transformer architectures to analyze user linguistic patterns and infer anxiety levels through interactive chat sessions. Unlike static conversational models, AegisMind fuses linguistic analysis with real-time biometric telemetry—specifically Blood Pressure, SpO_2, and heart rate variability—extracted via a wearable smartwatch device during live gaming or interaction blocks. When physiological stress markers spike, the wearable system deploys localized haptic feedback (vibrations) to disrupt pathological immersive loops. Furthermore, the system incorporates an operating-system-level continuous activity monitor that enforces hard-stop limits, automatically locking screens during intense gaming thresholds to mitigate cognitive overload. For critical cases, the framework leverages Retrieval-Augmented Generation (RAG) to serve as an empathetic conversational buffer while executing automated crisis-redressal handoffs to localized clinical practitioners. Experimental validations indicate that AegisMind bridges the critical gap between passive monitoring and immediate, ethical, automated psychological intervention.

Khetkart Smart Trading For Crops And Bids

Authors: Purnima Pandey, Aachal Patil, Siddhi J. Patil, Vipin Wani

Abstract: KhetKart will present a mobile application designed to connect farmers and buyers on a single digital platform. The application will enable farmers to list the crops, weeds, and e- waste they will have available, allowing buyers to browse and pur- chase items according to their needs. A built-in bidding system will let buyers place competitive offers on products, enhancing price transparency and negotiation. Additionally, the app will incorporate a geolocation feature that will display all buyers and sellers within a 50 km radius, promoting efficient and localized trade. By streamlining agricultural and e-waste exchange, the application will aim to reduce the role of intermediaries, support sustainable practices, and empower both farmers and buyers through direct, real-time interactions. Agricultural markets in India will continue to face challenges of price fluctuations, lack of transparency, and limited access to reliable information. Farmers and consumers will frequently depend on intermediaries for market prices, which will create inefficiencies and reduce profitability. To address this gap, we will develop Khet Kart, a mobile-based application that will provide a daily commodity price dashboard for local markets. The application will collect and organize real-time data of essential agricultural commodities and present it in an intuitive dashboard. By enabling farmers, traders, and consumers to make informed decisions, Khet Kart will ensure transparency, support fair pricing, and minimize exploitation by intermediaries. The system will not only benefit stakeholders at the grassroots level but will also contribute to strengthening the rural economy by integrating digital with agriculture.

DOI: http://doi.org/10.5281/zenodo.20929655

MID Meter PCB Soldering – Dual Heater

Authors: Atharv Nikam, Saurabh Jagtap, Priti Choudhari, Gaurav Uphade, Abhijit Lohakane

Abstract: Shunt soldering in MID (Measuring Instrument Directive) meter printed circuit boards (PCBs) is a critical manufacturing operation that directly affects electrical accuracy, reliability, and production throughput. Conventional manual soldering of multiple shunts is time-consuming and prone to inconsistency due to operator dependency. This paper presents the design and implementation of an automated dual-heater shunt soldering system for MID meter PCBs, capable of soldering four shunts per PCB with high precision and repeatability. The proposed system integrates a fixture-based automation mechanism controlled by a PLC, employing three pneumatic cylinders and a stepper motor–driven linear positioning system. One pneumatic cylinder securely clamps the PCB, while two heating cylinders equipped with soldering gun tips simultaneously heat both ends of the shunt to ensure uniform solder joints. The stepper motor indexes the fixture in four discrete steps of 17 mm, covering a total travel of 51 mm, followed by automatic homing. At each position, solder wire is applied and heated automatically. Experimental evaluation demonstrates significant improvements in cycle time, soldering precision, and process repeatability, making the system suitable for industrial MID meter production and scalable automation environments.

DOI: http://doi.org/10.5281/zenodo.20929743

Adaptive Clutter-Edge CFAR Detection For Pulsed Radar With Target Excision And Edge-Aware Threshold Selection

Authors: Kapil Dev Tyagi

Abstract: Constant false-alarm rate (CFAR) detection is fundamental to radar signal processing. Cell-averaging CFAR (CA-CFAR) is optimal in homogeneous Rayleigh clutter but suffers severe detection loss at clutter edges and in the presence of interfering targets. Order-statistic CFAR (OS-CFAR) is more robust at transitions but sacrifices detection sensitivity in uniform backgrounds. This paper proposes Adaptive Clutter-Edge CFAR (ACE-CFAR), a three-mode detector that (i) tests for clutter-power transitions using a leading-versus-lagging window ratio, (ii) selects the lower-power reference window at detected edges to prevent threshold inflation, and (iii) applies median-gated target excision in homogeneous regions to eliminate masking by nearby interferers. A soft sigmoid transition blends the two modes. The method is evaluated on synthetic Rayleigh-clutter range profiles with Swerling-I targets across four scenarios (homogeneous, clutter-edge, multi-target, and mixed) at seven input SNR levels using 200 Monte Carlo trials. ACE-CFAR more than doubles the probability of detection at clutter edges (Pd = 0.61 versus 0.25 for CA-CFAR and 0.20 for OS-CFAR) while maintaining competitive detection in homogeneous clutter (Pd = 0.71 versus 0.65). The measured false-alarm rate is modestly elevated (7-19 times the design Pfa) as an explicit trade-off for the detection gain, and is discussed openly. The results demonstrate that ACE-CFAR occupies a favourable operating point on the detection-versus-false-alarm surface that neither baseline reaches.

DOI: http://doi.org/10.5281/zenodo.20938103

Human-AI Skill Drift: Do AI Copilots Enhance Expertise Or Create Cognitive Dependency?

Authors: Rakesh Dondapati

Abstract: AI copilots are increasingly embedded in knowledge-intensive work environments, yet their long-term consequences for human expertise development remain inadequately theorized and empirically underexplored. This study introduces the concept of human-AI skill drift — defined as the gradual, often imperceptible change in worker cognitive capability and domain expertise resulting from sustained reliance on AI-assisted task performance. Drawing on a six-month longitudinal field experiment spanning 225 participants across three knowledge work contexts (consulting analysis, software development, and academic writing), employees were randomly assigned to five levels of AI copilot support: no assistance (control), minimal, balanced, full, and adaptive. The study measures productivity, output quality, independent task performance, error detection, learning retention, metacognitive accuracy, and AI over-reliance across three measurement waves. Analysis of covariance reveals that full AI support conditions produce the largest short-term productivity and quality gains (η²p = 0.46 and 0.41, respectively) but simultaneously generate the most severe skill drift (SDI = –0.61, p < .001) and the sharpest decline in independent task performance, error detection, and learning retention. Critically, the adaptive AI condition — in which AI assistance intensity is dynamically calibrated to real-time skill signals — achieves near-equivalent productivity and quality outcomes while preserving independent skill profiles comparable to minimal-assistance conditions. Qualitative analysis of 36 participant interviews yields five themes illuminating the experiential mechanisms of skill drift, including invisible skill erosion, calibration failure, and accountability diffusion. The study contributes a validated Skill Drift Index (SDI), an AI-mediated learning theory, and a Responsible Augmentation Design Framework to the human-computer interaction and future-of-work literatures.

DOI: http://doi.org/10.5281/zenodo.20959526

A Review Of Deep Learning Architectures For Multi-Class Skin Lesion Classification: From CNN-Recurrent Hybrids To Class-Balanced Peephole LSTM Frameworks

Authors: Akash saini, Jitender Kumar

Abstract: Automated dermoscopic image analysis has emerged as a critical tool for addressing the global burden of skin cancer, yet two persistent obstacles continue to limit clinical translation: high inter-class morphological similarity among diagnostic categories and severe class imbalance within benchmark datasets such as HAM10000. This paper reviews the methodological evolution of deep learning approaches for multi-class skin lesion classification, tracing the progression from handcrafted feature-based classical machine learning pipelines through deep convolutional architectures, hybrid CNN-recurrent frameworks, and Vision Transformer-based models. Particular attention is given to the gate desynchronisation limitation inherent in standard Long Short-Term Memory units, which excludes the internal cell state from gate computations and can cause premature loss of diagnostically relevant sequential information. The review further examines loss-level and data-level strategies for class-imbalance mitigation, including Weighted Cross-Entropy, Focal Loss, SMOTE, and generative adversarial augmentation, and surveys the explainability frameworks required for regulatory and clinical acceptance. Based on the gaps identified, the paper discusses an emerging direction—class-balanced hybrid CNN-Peephole LSTM frameworks—and outlines future research priorities, including attention-based feature filtering, multi-modal metadata fusion, and Vision Transformer knowledge distillation for resource-constrained clinical deployment.

DOI: http://doi.org/10.5281/zenodo.20964288

A Comprehensive Review on the Design and Performance of High-Speed External Gear Pump for Aerospace Applications

Authors: Sameer Ahmed Khan, Dr. P Rama Lakshmi

Abstract: High-speed external gear pumps play a crucial role in aerospace systems, particularly in lubrication, fuel delivery, and hydraulic actuation systems. Their compact size, reliability, and ability to operate at high rotational speeds make them highly suitable for demanding aerospace environments. This paper presents a comprehensive review of the design aspects and performance characteristics of high-speed external gear pumps. The study focuses on key parameters such as gear geometry, material selection, clearance optimization, and operating conditions that significantly affect pump efficiency and durability. Additionally, the impact of leakage losses, cavitation, thermal effects, and fluid properties on performance is discussed. Advanced computational tools such as Computational Fluid Dynamics (CFD) and Finite Element Analysis (FEA) have enabled better understanding of internal flow and structural behavior. Challenges such as noise, vibration, and pressure fluctuations are also addressed. The paper concludes with future research directions including additive manufacturing, smart monitoring systems, and advanced coatings for improved performance.

A Review On Design Considerations And Alignment Techniques For Extramedullary Tibial Jigs In Distal Femoral Reconstruction

Authors: Shaik Farhaan, Dr. Ch. Indira Priyadarshini

Abstract: Distal femoral reconstruction has become one of the most effective limb-salvage procedures for patients suffering from malignant bone tumors, severe trauma, revision arthroplasty, and extensive bone defects. The long-term clinical success of these procedures depends largely on accurate tibial alignment, which directly influences implant positioning, joint biomechanics, wear characteristics, prosthesis longevity, and postoperative rehabilitation. Conventional intramedullary alignment systems provide satisfactory accuracy during routine total knee arthroplasty; however, their application is limited in cases involving tumor resection, previous implants, deformities, or obstruction of the medullary canal. Consequently, extramedullary tibial alignment systems have emerged as an effective alternative because they utilize external anatomical landmarks without violating the medullary canal. Recent developments in computer-aided design (CAD), finite element analysis (FEA), patient-specific instrumentation, additive manufacturing, and biomechanical optimization have enabled the development of highly accurate and surgeon-friendly extramedullary tibial jigs. These technological improvements contribute to enhanced alignment precision, reduced surgical complexity, and improved implant survival rates. This review summarizes the current state of research on extramedullary tibial alignment systems, emphasizing biomechanical principles, engineering design considerations, material selection, CAD-based development, finite element validation, manufacturing technologies, and future research directions. Furthermore, existing challenges and future opportunities in orthopedic instrumentation are critically reviewed. The review highlights the importance of integrating engineering design with clinical requirements to improve surgical outcomes and support the development of next-generation orthopedic alignment systems.

AI-Powered Interior Design System

Authors: Hemant Sharma, Ankit Rajak, Dr Jasbir Kaur, Assistant Professor Ifrah Kampoo, Assistant Professor Mansi Rajapurkar

Abstract: This research paper presents an AI-powered interior design system developed using Generative Artificial Intelligence techniques for automated room visualization, smart furniture arrangement, and personalized interior recommendations. Traditional interior designing methods often require significant time, professional expertise, and high financial investment, making them less accessible for common users. The proposed system aims to simplify and modernize the interior designing process by integrating advanced AI models capable of generating realistic and customized room designs based on user preferences, room dimensions, color combinations, and design themes. The system utilizes prompt-based image generation models such as diffusion models and deep learning techniques to generate visually appealing interior layouts in real time. Users can provide textual prompts describing their preferred room style, including themes such as modern, minimalist, luxury, vintage, or contemporary, and the AI system generates corresponding room visualizations automatically. Experimental analysis demonstrates that the AI-based system significantly reduces interior planning time while improving visualization quality and user satisfaction when compared with traditional manual design approaches.

DOI: https://doi.org/10.5281/zenodo.21022452

Smart Attendance Management System Using Face Recognition, Liveness Detection, and Geolocation Verification

Authors: Omkar Mane, Dr.Jasbir Kaur, Assistant Professor Mansi Rajapurkar

Abstract: Conventional attendance systems can be quite time-consuming and are often vulnerable to proxy attendance. This paper presents a Smart Attendance Management System that leverages face recognition, prompt-based liveness detection, and geofencing to streamline and secure the attendance process. Built on a web-based application platform using Python Flask, OpenCV, HTML, CSS, JavaScript, and MySQL, the system starts by capturing a student's face through a webcam, then employs face recognition to verify their identity. To prevent the use of fake photos or videos, it incorporates a prompt-based liveness detection feature, where students must read prompts that appear randomly on the screen. Additionally, attendance is only recorded if the student is physically present in the classroom, thanks to GPS-based geofencing. The system also provides an analytics dashboard for reviewing attendance records and trends. This proposed solution not only enhances accuracy but also significantly reduces manual effort, offering a substantial boost in security and reliability for attendance management in educational institutions.

DOI: https://doi.org/10.5281/zenodo.21023019

Real-Time Intelligent Waste Classification and Environmental Impact Feedback Using YOLOv8

Authors: Atharva Subhash Patil, Pritish Narayan Patil, Dr. Jasbir Kaur, Ifrah Kampoo, Mansi Rajapurkar

Abstract: The rapid growth of urban waste has created a major challenge for environmental sustainability and waste management operations, especially in cities where segregation is performed manually and often inconsistently. Manual sorting is slow and vulnerable to human error, contamination, and operational inefficiency, particularly when multiple waste categories are mixed together. To address this issue, this paper presents a real-time waste classification framework based on YOLOv8, combined with environmental impact feedback through decomposition-time information. The model was trained using a mixture of synthetic and real waste images covering six waste classes: plastic, paper, metal, glass, organic waste, and cardboard. YOLOv8 was selected because it is designed for real-time object detection and supports standard evaluation using precision, recall, and mean Average Precision metrics. The model was further fine-tuned on real-world waste images to improve robustness in practical conditions such as clutter, illumination changes, partial occlusion, and background variation. Experimental results showed a validation mAP@0.5 of approximately 99.4%, with precision and recall near 99%. In addition to detection, the system presents decomposition-time and disposal guidance, which helps increase environmental awareness and encourages responsible waste handling.

DOI: https://doi.org/10.5281/zenodo.21023586

A Review On Magnesium-Based Metal Hydride Reactors For Efficient Solid-State Hydrogen Storage: Materials, Thermal Management And Design Approaches

Authors: Rama Satya Sai, Dr. Rahul

Abstract: The global transition toward sustainable energy systems has accelerated interest in hydrogen as a clean and renewable energy carrier. However, the widespread adoption of hydrogen technologies is strongly dependent on the development of efficient, safe, and economically viable hydrogen storage methods. Among the available storage techniques, solid-state hydrogen storage using metal hydrides has emerged as one of the most promising approaches because of its high volumetric hydrogen density, enhanced operational safety, and reversible hydrogen absorption–desorption characteristics. Magnesium-based metal hydrides (MgH₂) have attracted considerable attention owing to their high theoretical hydrogen storage capacity (7.6 wt.%), natural abundance, lightweight nature, and relatively low cost. Despite these advantages, practical implementation remains challenging due to slow hydrogen sorption kinetics, poor thermal conductivity, and high hydrogen desorption temperatures. This review presents a comprehensive overview of recent developments in magnesium-based metal hydride reactor technology. Various hydrogen storage techniques are compared, followed by an in-depth discussion of hydrogen absorption mechanisms, reactor configurations, heat transfer enhancement methods, catalyst-assisted performance improvement, and advanced reactor design strategies. Recent progress in thermal management techniques, including metallic fins, expanded graphite, metal foams, heat exchanger tubes, and nanostructured additives, is critically reviewed. Furthermore, current research gaps and future opportunities in reactor optimization, additive manufacturing, artificial intelligence-based design optimization, and renewable energy integration are discussed. This review aims to provide researchers with a comprehensive understanding of the current state of magnesium-based hydrogen storage technology and its potential contribution to future hydrogen energy systems.

An Intelligent Smart Ambulance Framework Using IoT And Artificial Intelligence For Real-Time Healthcare Coordination

Authors: Dr. Pankaj Malik, Ridima Sharma, Mansi Dubey, Manya Gautam, Vinayak Mishra

Abstract: Timely emergency medical care plays a crucial role in reducing mortality and improving patient outcomes. Conventional ambulance systems often suffer from delayed hospital communication, inefficient route selection, and limited patient monitoring during transportation. This paper presents an IoT-Enabled Smart Ambulance with Real-Time Hospital Communication Using Artificial Intelligence (AI) to enhance emergency healthcare services. The proposed system integrates IoT-based medical sensors, GPS tracking, cloud computing, and AI-driven analytics to continuously monitor patient vital parameters such as heart rate, blood pressure, oxygen saturation (SpO₂), and body temperature. These data are transmitted in real time to hospitals, enabling medical personnel to assess patient conditions and prepare appropriate treatment before arrival. The AI module performs patient severity classification, estimated time of arrival (ETA) prediction, hospital recommendation, and intelligent route optimization based on real-time traffic conditions. A cloud-based communication platform facilitates seamless information exchange between ambulance staff and healthcare providers. The system was evaluated using simulated emergency healthcare datasets and real-time traffic information. Experimental results demonstrate that the proposed framework reduces ambulance travel time by 28.4%, improves ETA prediction accuracy to 95.6%, and decreases hospital preparation delays by 41.7% compared with conventional ambulance systems. Furthermore, patient condition classification achieved an accuracy of 97.2%, enabling faster medical decision-making. The integrated IoT-AI architecture significantly improves emergency response efficiency, resource utilization, and patient survival probability. The proposed smart ambulance framework provides a scalable and intelligent solution for next-generation emergency healthcare systems and smart city applications.

DOI: http://doi.org/10.5281/zenodo.21028015

A Comprehensive Review of Deep Learning Techniques for Brain Tumor Detection and Classification Using MRI Images

Authors: Jyoti Gahora, Bhanu Pratap Singh

Abstract: Brain tumor detection and classification are among the most critical challenges in medical image analysis due to the complexity and variability of tumor structures in Magnetic Resonance Imaging (MRI) scans. Early and accurate diagnosis is essential for effective treatment planning and improving patient survival rates. In recent years, deep learning techniques have significantly transformed the field of medical imaging by providing automated, efficient, and highly accurate diagnostic systems. This review paper presents a comprehensive analysis of recent advancements in machine learning and deep learning approaches for brain tumor detection and classification using MRI images. The study examines various Convolutional Neural Network (CNN) architectures, transfer learning models, attention mechanisms, hybrid frameworks, explainable artificial intelligence (XAI), and IoT-enabled healthcare systems. Additionally, the paper discusses preprocessing methods, segmentation techniques, classification strategies, and performance evaluation metrics used in recent research. The review also identifies major challenges such as limited annotated datasets, computational complexity, overfitting, lack of interpretability, and generalization issues across medical datasets. Finally, the paper highlights emerging trends and future research directions, including lightweight deep learning models, federated learning, multimodal imaging integration, and real-time clinical deployment. This review provides researchers and healthcare professionals with a detailed understanding of state-of-the-art deep learning techniques for intelligent brain tumor diagnosis systems.

An Intelligent Deep Learning Framework for Violence Detection and Criminal Activity Identification in Smart Surveillance Systems

Authors: Shivam Namdev, Bhanu Pratap Singh

Abstract: The increasing growth of urbanization, public gatherings, and security threats has created a strong demand for intelligent surveillance systems capable of automatically detecting violent activities and identifying criminal behavior in real time. Traditional surveillance systems mainly rely on manual monitoring, which often leads to delayed response, human errors, and inefficient threat analysis. To address these limitations, this research proposes an Intelligent Deep Learning Framework for Violence Detection and Criminal Activity Identification in Smart Surveillance Systems using advanced Deep Learning architectures and video analytics techniques. The proposed framework integrates preprocessing, segmentation, feature extraction, and hybrid Deep Learning models including ResNet50, MobileNetV2, LSTM, and the proposed InceptionV3 + LSTM architecture for violence classification. The system processes surveillance videos by extracting spatial and temporal features to accurately distinguish violent and non-violent activities. The dataset consists of two categories, namely Violence and Non-Violence, containing real-world surveillance scenarios such as sports, crowd movement, eating, singing, and violent human interactions. Experimental analysis demonstrates that the proposed InceptionV3 + LSTM model achieved superior performance with improved training accuracy, validation accuracy, and reduced loss values compared with existing models. The framework effectively captures complex motion patterns, human interactions, and abnormal behaviors in surveillance videos while reducing false detections. The proposed intelligent surveillance system can be applied in smart cities, public transportation, educational institutions, stadiums, and crowded environments to improve public safety, automated threat detection, and real-time security monitoring.

Integration of Iot Temperature Sensors in Plastic-Perlite Bricks for Building Energy Monitoring

Authors: Rasika Appaso Kharat, Pratiksha Sanjay Mane, Parvej Najirhusen Shaikh, Dr. S. B. Walke

Abstract: The rapid growth of urban infrastructure has increased the demand for energy-efficient and sustainable construction materials. This study focuses on the integration of IoT-based temperature sensors within plastic-perlite composite bricks to enhance building energy monitoring and management. The proposed system utilizes waste plastic as a partial replacement material combined with expanded perlite to develop lightweight, thermally efficient, and eco-friendly building units compatible with standard M20-grade concrete structures. IoT temperature sensors (DS18B20) are embedded within the composite bricks to continuously monitor internal and external temperature variations, enabling real-time data collection and analysis through wireless communication modules. Laboratory tests, including compressive strength, thermal conductivity, and water absorption, are conducted to evaluate the structural and thermal behavior of the developed bricks. The experimental results show that the integration of perlite and waste plastic significantly improves thermal resistance, reduces heat transfer, and decreases structural dead load by lowering the average brick density to 821–907 kg/m³ compared to 1800 kg/m³ for conventional units. This research promotes the development of sustainable smart construction materials, aligning with modern green building technologies and the global agenda for energy efficiency.

Structural Performance Of High-Rise Buildings With Floating Columns For Long-Span Applications: A Comprehensive Review

Authors: Madhusudan Dubey, Murlidhar Chourasia, Dr. Rahul Kumar Satbhaiya

Abstract: This review paper provides a synthesis of the experimental, analytical and computational studies that have been published in the last 30 years concerning floating-column structural systems with particular emphasis on seismic performance, transfer-beam performance, compliance with the IS-codes and economic considerations.Laboratory studies and finite-element studies have shown that floating columns can produce lateral displacements 80 to 101% greater than equivalent regular columns; bending moments in the transfer columns can be increased by 200-362% due to the presence of floating columns, and design base shear can be increased by 20-30% due to floating columns compared to equivalent regular columns. In Floating Column building the bending moment of the transfer beam is in the ratio of 4-7 compared to equivalent regular beam, which needs special detailing of deep beams as per IS 456: 2000 and IS 13920: 2016. The structural cost premium is 14-20% as per zone and height of building. Research areas include non-linear time-history analysis, integration with base-isolation, machine-learning and transfer system optimisation and performance-based seismic design (PSD) in higher seismic zones (PSD III-V).

AI Driven Strategic Decision Support System Using Hybrid Artificial Intelligence And Particle Swarm Optimization

Authors: Samyo Ranjan Jagdev, Subhashree Sibani Sahu, Achinta Kumar Palit, Sumant Sekhar Mohanty

Abstract: Strategic decision-making has become an increasingly important function in today’s corporate world. The challenge for managers is to evaluate many options where there are unknown factors (e.g., the condition of the market, financial constraints, and competitors). Managers currently rely heavily on experience and qualitative methods for making strategic decisions. Traditional methods do not take full advantage of the vast volume of data generated by the modern enterprise. With the emergence of Artificial Intelligence (AI), organizations can analyze large-scale databases to generate predictive insight for making data-driven decisions. This paper proposes a hybrid AI-based strategic decision support system. The system combines machine learning and Particle Swarm Optimization (PSO) to forecast strategic decisions and optimize strategic decision-making variables. The paper outlines the mathematical modelling; the overall design of the system, algorithms, and experimental evaluation results. Results from the experiments showed that the hybrid AI+PSO framework improved the accuracy of the decision and effectively identified optimal strategies when compared to traditional machine learning approaches

Bridging the Learning Gap: An Integrated Solution for Notes, PYQs, Live Classes, and AI Tools

Authors: Ms. Mrunmayi Ashok Kurhade, Dr. Jasbir Kaur, Assistant Professor Ms. Mansi Rajapurkar

Abstract: The modern learning environment is characterized by fragmentation, where students often rely on multiple disconnected platforms for accessing notes, previous year questions (PYQs), live classes, and AI-driven tools. This disjointed approach results in inefficiency, reduced engagement, and difficulty in tracking progress. To address these challenges, this paper proposes an integrated solution that unifies all essential learning resources into a single platform. The system architecture comprises modules for notes repository, PYQ management, AI recommendation engine, live class integration, and analytics. Artificial intelligence techniques such as natural language processing (NLP), collaborative filtering, reinforcement learning, and adaptive analytics are employed to personalize learning, resolve doubts, and track student progress. The proposed solution not only enhances accessibility and efficiency but also lays the foundation for future integration of immersive technologies such as AR/VR and predictive analytics.

DOI: https://doi.org/10.5281/zenodo.21061973

Cybersecurity in Modern Digital Ecosystems: Threat Detection, Prevention, and Resilient Architecture

Authors: Rohan Tiwari, Sanket Jadhav, Dr. Jasbir Kaur, Assistant Professor Suraj Kanal, Assistant Professor Ifrah Kampoo

Abstract: The rapid proliferation of cloud computing, artificial intelligence, the Internet of Things (IoT), and distributed digital services has fundamentally altered the modern computing environment. While these advancements improve operational efficiency and connectivity, they also introduce sophisticated security vulnerabilities that frequently bypass traditional defensive mechanisms. Contemporary cyber threats have evolved from simple malicious software into complex, targeted campaigns, including advanced persistent threats (APTs), zero-day exploits, ransomware, phishing, and internal vulnerabilities. As a result, conventional perimeter-based security strategies are no longer sufficient for protecting increasingly expansive digital infrastructures. This research examines the changing landscape of cyber threats and assesses the limitations of legacy security frameworks in protecting modern digital ecosystems. The paper explores innovative developments in machine learning-driven intrusion detection, blockchain-based security, intelligent threat recognition, and Zero Trust Architecture (ZTA). Building on these insights, the study proposes a multi-layered cybersecurity framework that incorporates behavioral analytics, continuous monitoring, robust encryption, artificial intelligence, and automated response mechanisms. The goal of this proposed architecture is to strengthen organizational resilience against both known and emerging threats by improving detection accuracy and reducing the time required for threat mitigation in scalable, cloud-native environments.

DOI: https://doi.org/10.5281/zenodo.21062623

CleanTalk: An NLP-Based Framework for Profanity Detection and Severity Analysis in Video Content

Authors: Om Agarwal, Dr. Jasbir Kaur, Assistant Professor Ifrah Kampoo

Abstract: This paper presents CleanTalk, a Natural Language Processing (NLP)-driven system designed to automate the detection and severity assessment of profane language within video content. The system operates by extracting audio tracks from video files, transcribing the resulting audio into textual form using speech-to-text technology, and subsequently applying NLP-based analytical techniques to identify profane words and evaluate their severity. Each detected instance is annotated with a corresponding timestamp, and the system produces an overall severity score summarizing the nature of the language throughout the content. The final output consists of an annotated video file with visual markers at profanity timestamps alongside a structured severity report. A pilot evaluation was conducted across three video samples of varying content type, yielding an average precision of 96.7% and an average recall of 84.5%, with performance degradation observed primarily in multi-speaker and fast-speech scenarios. CleanTalk demonstrates strong potential as a scalable, automated solution for content moderation workflows on digital platforms.

DOI: https://doi.org/10.5281/zenodo.21065156

Pesticide Residue contamination and Health risk assessment in Channa sp. from Degganahalli, Mysore

Authors: Research Scholar Kavitha, Krishna M P

Abstract: Water pollution through pesticides residues and their accumulation in different tissue of aquatic animals and crops has become a real threat to human and other animals on earth. OCP’s are found cause health risks including cancer. The current study focuses on assessing the Organochloride contamination in the fish tissue (Channa.sp.) from Cauvery River, near Degganahalli area of Mysore district. Fish samples were collected from the sample site and processed following FSSAI test method using GCMS/MS. Hazard Quotient was calculated for a sample size of 384 using EDI and ADI. Concentration levels of pesticides were in the order Dieldrin> Aldrin> γ-HCH>β HCH. Corresponding concentrations of pesticides are 7.73 ± 0.94 > 1.07 ± 0.15> 0.08 ± 0.03 > 0.06 ± 0.03 respectively. Hazard quotient of dieldrin is significant with a Mean± SD value of 0.05 ± 0.03 followed by Aldrin,0.00662 ± 0.00441, γ-HCH, 0.00050 ± 0.00033 and β-HCH, 0.00037 ± 0.00025 respectively. Pesticide levels in the fish are found to be below permissible limits across all seasons od study period. Hazard quotient also falls under low risk level. Detection of Aldrin and dieldrin in the current study signifies the presences of POPs in the environment.

DOI: https://doi.org/10.5281/zenodo.21065820

Blood Fruit, Haematocarpus Validus Bakh. F. Ex Forman: A Underutilized Edible Plant From India

Authors: Rekha Bora

Abstract: Blood fruit (Haematocarpus validus Bakh. f. ex Forman) is a lesser-known, underutilized woody climber or liana belonging to the family Menispermaceae, mainly found in Northeast India, the Andaman and Nicobar Islands, and few other parts of Southeast Asia. It has long been used by indigenous communities as a fruit and traditional medicine for treating blood-related disorders, jaundice, and skin ailments. The fruit is rich in iron, vitamin C, carotenoids, phenolic compounds, flavonoids, and anthocyanins, making it a valuable source of natural antioxidants with potential applications in the food, pharmaceutical, nutraceutical, and natural dye industries. This review summarizes the current knowledge on the taxonomy, distribution, botanical characteristics, traditional uses, nutritional and phytochemical composition, pharmacological properties, ecological importance, and commercial potential of H. validus. It also focusses on current work on its anthocyanin content, physicochemical properties, and value-added applications. Although recent studies have been conducted on this plant, research on this species remains limited. Therefore, this review identifies the major research gaps and emphasizes the need for further studies on conservation, cultivation, product development, and sustainable utilization to promote the wider use of this valuable native fruit while supporting biodiversity conservation and rural livelihoods.

DOI: http://doi.org/10.5281/zenodo.21065874

YOLOv5 and EfficientDet based Object Detection and Classification for Visually Impaired

Authors: Khushi Sinha, Esha Garg, Gaurav Singal, Preeti Kaur

Abstract: This research presents a novel approach that utilizes the EfficientDet model to address the challenges faced by people with visual impairments in recognizing their environment. The project focuses on the integration of state-of-the-art computer vision technology into assistive devices, specifically blind sticks, to improve the perception of the environment. The EfficientDet model, known for its superior efficiency- accuracy tradeoff, is trained to detect and classify different objects in real time providing valuable insights to users. The project contributes to the field of assistive technology by using state-of-the-art machine learning models to improve accessibility and autonomy for people with visual impairments.

DOI: https://doi.org/10.5281/zenodo.21067362

Takayasu Arteritis In Children And Young Adults: Current Perspectives On Pathogenesis, Clinical Features, Diagnosis, Monitoring And Management

Authors: Preethi G N, Ashwini Angadi, Janaki R Torvi, Shaista Omer, Adarsh GS

Abstract: Takayasu arteritis (TA) is a rare chronic granulomatous vasculitis involving the aorta and its major branches. The disease predominantly affects young females and may lead to progressive vascular stenosis, occlusion, aneurysm formation, and ischemic complications. Early diagnosis remains challenging because initial symptoms are often nonspecific and laboratory markers may not accurately reflect disease activity. Advances in imaging techniques, immunosuppressive therapy, biologic agents, and vascular interventions have significantly improved patient outcomes. This review summarizes the epidemiology, pathogenesis, clinical manifestations, diagnosis, disease monitoring, treatment strategies, and prognosis of Takayasu arteritis, with special emphasis on pediatric disease.

Solid Waste Management in Urban India: Issues, Challenges, and Sustainable Solutions

Authors: Assistant Professor Lalit Saikia

Abstract: In India with high urbanization and rapid population growth, solid waste management is a matter of concern due to generation of large amount of waste, improper management and associated problems of environmental pollution, health issues and economic impacts. Based on online survey among different groups of people from major towns and cities of India and review of literature, this paper attempts to highlight issues, concerns and promising options for solid waste management in urban India. It is revealed that public awareness on reduce and reuse, waste segregation at source, available dustbins in proper places, sufficient man-power with equipments for regular collection and disposal with advanced technologies and strict rules are need of the hour. Generated waste can be converted into a potential resource if treated properly with effective waste management technologies.

DOI: https://doi.org/10.5281/zenodo.21097136

Design and Development of a Sales Performance Dashboard for the BFSI Sector Using Advanced Data Visualization Tools

Authors: Rahul Praful Gaonkar, Shubham Kashinath Jadhav, Dr. Jasbir Kaur, Assistant Professor Suraj Kanal, Assistant Professor Sandhya Thakkar

Abstract: The Banking, Financial Services, and Insurance (BFSI) sector operates within a highly competitive and data-heavy ecosystem, generating vast amounts of transactional, customer, and sales data daily. However, extracting actionable insights from this fragmented, multi-source raw data to drive sales performance and strategic decision-making remains a critical challenge. This paper outlines the design, architectural framework, and implementation of a comprehensive Sales Perfor-mance Dashboard tailored specifically for the BFSI sector using Power BI. The novelty of this work lies in the formulation of an optimized, high-throughput Extract, Transform, and Load (ETL) pipeline linked to a custom Star Schema data model that successfully blends disparate core banking ledgers, CRM logs, and insurance pipelines into a unified repository. Operat-ing on a multi-million-row financial dataset, the key outcomes demonstrate an operational reporting time latency reduction of 98.6% (from 3–5 days down to under 5 minutes) alongside a rapid rendering processing response under 1.2 seconds. The im-plementation demonstrates how interactive analytics eliminates information silos and empowers financial supervisors with real-time operational visibility.

DOI: https://doi.org/10.5281/zenodo.21099318

Criminal Face Detection

Authors: Khatik Huzaifa Mohammad Aarif, Dr. Jasbir Kaur, Professor Sandhya Thakkar

Abstract: Criminal face detection is a critical aspect of law enforcement and public safety. This project explores the application of machine learning and computer vision techniques to identify potential criminal faces from images. The methodology involves preprocessing image data, extracting facial features using deep learning models like Convolutional Neural Networks (CNNs), and implementing facial recognition algorithms .The project utilizes popular Python libraries such as Open CV , Tensor Flow, and Keras to train and deploy the models. Additionally, a dataset comprising diverse facial images is employed for model training and evaluation. The trained model's performance is assessed using metrics such as accuracy, precision, recall, and F1 score. Results demonstrate the feasibility of using machine learning algorithms to detect potential criminal faces with a certain degree of accuracy. Ethical considerations regarding biases in data and implications of using such technology in law enforcement are discussed. Further research directions are suggested to enhance the robustness and fairness of criminal face detection systems. This abstract provides a high-level overview of the project's objectives, methodologies, findings, and potential ethical considerations without getting into specific code implementations or technical details.

DOI: https://doi.org/10.5281/zenodo.21161174

A Lightweight, Generalizable Machine Learning Framework for Formula 1 Race Prediction: A Las Vegas Grand Prix Case Study

Authors: Sahil Jeetendra Vyas, Soham Krishna Morlikar, Dr. Jasbir Kaur, Assistant Professor Suraj Kanal, Assistant Professor Sandhya Thakkar

Abstract: Formula 1 is one of the most technologically advanced motorsports in the world, generating enormous volumes of performance data throughout every race weekend. Accurate prediction of race outcomes remains a challenging task because race performance is influenced by numerous dynamic factors including qualifying pace, tyre degradation, weather conditions, pit-stop strategy, mechanical reliability, safety car deployments, and driver skill. Since most telemetry collected by Formula 1 teams is proprietary, researchers must rely on publicly available datasets to develop reproducible prediction models. This paper presents a lightweight machine learning framework for predicting Formula 1 race finishing times and projected finishing positions using publicly available qualifying data obtained through the FastF1 API. Unlike existing approaches that require extensive telemetry or large feature sets, the proposed framework intentionally utilizes qualifying lap time as the primary predictive feature in order to investigate its standalone predictive capability. A Gradient Boosting Regressor is employed to estimate race finishing times after preprocessing qualifying and race timing data into numerical representations. Predicted race times are subsequently ranked to generate projected finishing positions. Model performance is evaluated using Mean Absolute Error (MAE), while an interactive Streamlit dashboard provides an intuitive visualization interface for exploring prediction results and driver rankings. Experimental evaluation demonstrates that qualifying performance contains substantial predictive information regarding race outcomes while maintaining low computational complexity and complete repro-ducibility using publicly accessible data. The proposed methodology establishes a transparent baseline framework for future Formula 1 race prediction research and provides a foundation for incorporating additional race variables in future studies.

DOI: https://doi.org/10.5281/zenodo.21104732

From Clinical Interview To Computational Signal: A Review Of DAIC-WOZ-Based Depression Detection Systems

Authors: Harsh, Dr. Pramod Kumar

Abstract: Major Depressive Disorder (MDD) is a disease that impacts more than 280 million people world-wide, and is woefully underdiagnosed because of subjectivity of clinical interview-based assessment measures. Distress Analysis Interview Corpus – Wizard of Oz (DAIC-WOZ) has become the key baseline dataset to assess automated, multimodal computational systems for depression screening. This review systemically discusses the machine learning and deep learning architectures tested on DAIC-WOZ ranging from unimodal facial, acoustic and textual approaches up to cutting-edge multimodal fusion frameworks that involve cross-modal attention and temporal graph neural networks. Important methodological weaknesses, subject leakage and shortcut learning through therapists prompts, are tested and quantified. Results show that some results of multimodal systems have a binary F1 score greater than 0.91, although many of the results are extremely overrated due to obsolete implementation of data division. Finally, the methodological recommendations and directions for future research to create clinically deployable, privacy-preserving and demographically generalizable depression screening systems are provided.

DOI: http://doi.org/10.5281/zenodo.21126640

D4PG-Based Energy-Aware Client Selection for Federated Learning in Industrial IoT

Authors: Dr. M. Mahil, S. Sindhu

Abstract: Industrial Internet of Things (IIoT) environments consist of a large number of heterogeneous edge devices with varying computational capabilities, energy levels, and communication conditions. Traditional Federated Learning (FL) approaches typically employ random or static client selection strategies, which often lead to excessive energy consumption, increased communication overhead, and slow model convergence. To address these challenges, this paper proposes a D4PG-based Energy-Aware Client Selection framework for Federated Learning in Industrial IoT systems. The proposed approach models client participation as a sequential decision-making problem and utilizes Deep Distributed Distributional Deterministic Policy Gradient (D4PG) reinforcement learning to dynamically select clients based on battery level, CPU utilization, bandwidth availability, latency, and expected training contribution. A multi-objective reward function is designed to jointly optimize learning accuracy, energy efficiency, and communication cost. The framework is evaluated in a heterogeneous non-IID federated environment consisting of 20–50 edge clients. Experimental results demonstrate that the proposed D4PG-based strategy improves convergence speed, reduces communication overhead, and lowers energy consumption compared with conventional client select Federated Learning ion methods. The findings highlight the effectiveness of distributional reinforcement learning for intelligent resource-aware federated optimization in Industrial IoT deployments.

DOI: https://doi.org/10.5281/zenodo.21126888

IV Planner: A Web Platform for Educational Visit Planning, QR Attendance, Geofencing, and Emergency Alerts

Authors: Omkar Gawde, Nupun Sawant, Dr. Jasbir Kaur, Assistant Professor Mansi Rajapurkar

Abstract: Organizing institutional educational visits involves a complex set of tasks itinerary coordination, attendance verification, real-time safety oversight, and post-trip documentation that most colleges still handle through paper registers, printed schedules, WhatsApp groups, and phone calls. This paper describes IV Planner, a web-based platform purpose-built for managing educational field visits. The platform brings together AI-assisted itinerary drafting (powered by Google Gemini with JSON schema validation), QR-code attendance, GPS-aware geofencing using the Haversine formula, GPS-tagged photo logging, and Socket.IO-driven emergency alerts under one unified interface. The backend stack comprises React/Vite on the client side, Express/Node.js on the server, MongoDB for persistence, and a structured validation layer that treats every Gemini response as untrusted until it clears required-field checks and server-side range validation. A 30-day field deployment with 52 participants across four institutional visits measured SOS dispatch latency, geofence notification time, QR scan confirmation, API response times, WebSocket propagation delay, battery consumption, and concurrent connection stability. Manual baselines were observed separately and converted to milliseconds purely for operational comparison. The deployment recorded a mean SOS dispatch latency of 310 ± 42 ms, geofence notification latency of 480 ± 65 ms, and QR scan-to-confirm time of 620 ± 80 ms. The system handled 350 simultaneous WebSocket connections without errors and was stress-tested to 500, where a 1.24% error rate appeared. Findings indicate that combining WebSocket event delivery, server-side geofence validation, database-enforced attendance integrity, and structured LLM output validation is feasible for real-world educational visit scenarios, though broader multi-institution studies are still needed before firm generalizations can be drawn.

DOI: https://doi.org/10.5281/zenodo.21157813

Interactive Olympic Data Analysis and Visualization Using Python and Streamlit

Authors: Omkar Jadhav, Kaustubh Naik, Dr. Jasbir Kaur, Assistant Professor Mansi Rajapurkar

Abstract: The exponential growth of sports data has created a compelling opportunity to extract actionable insights through interactive analytical platforms. This paper presents the design and implementation of a web-based Olympic data analysis system developed using Python and Streamlit. The proposed system processes a comprehensive historical dataset spanning 120 years of Olympic competition, encompassing more than 271,000 athlete-event records. Analytical modules are developed to examine medal tallies, country-wise performance trajectories, athlete biometric distributions, sport-wise participation patterns, and longitudinal gender participation trends. Data preprocessing employs Pandas-based pipelines to address missing values, eliminate duplicate records, and engineer derived features. Visualization is achieved through a multi-library strategy utilizing Matplotlib for static charts, Seaborn for statistical graphics, and Plotly for fully interactive, user-driven figures. The Streamlit framework provides a reactive web interface enabling real-time filtering without requiring server-side scripting expertise. Benchmark measurements indicate an average dashboard load time of 1.8 seconds and a full-dataset processing time of 0.42 seconds. The study identifies limitations in real-time data integration and proposes future extensions including machine-learning-based medal prediction and cloud deployment.

DOI: https://doi.org/10.5281/zenodo.21129423

Real-Time Position Estimation of a Variable Reluctance Resolver Using AD2S1205 and TI F28388D Through SPI-Based Adaptive Digital Processing

Authors: Ravikiran R. N, Dr.R.T Ugale, Srikant Warpe

Abstract: Variable reluctance resolvers are extensively used in traction drives, industrial automation, and aerospace applications due to their robustness and reliable operation under harsh environmental conditions. The quality of rotor position feedback directly affects the performance of vector-controlled electric machines. This work presents the implementation of a real-time VR resolver interface using the AD2S1205 resolver-to-digital converter and the TI-F28388D microcontroller connected through a high-speed Serial Peripheral Interface (SPI). The proposed embedded architecture continuously acquires angular position and velocity information and performs digital post-processing to improve measurement quality. In addition to resolver data acquisition, low-pass and adaptive filtering techniques are investigated to suppress electrical noise and reduce estimation jitter across different rotational speeds. Angle unwrapping and derivative-based speed computation are incorporated to ensure continuous position tracking under dynamic operating conditions. The proposed methodology is intended for applications requiring high-resolution feedback such as electric-vehicle drives and servo systems. Simulation and hardware validation are planned to evaluate position error, speed ripple, communication latency, and computational overhead. The presented approach provides an efficient and scalable framework for integrating resolver sensors with modern embedded motor control platforms.

Design and Fabrication of a Simulator for the Fire Detection and Fire Extinguishing System on the MI-8 Helicopter

Authors: MSc Trong Son Phan, MSc Van Huong Ngo, MSc Le Phan

Abstract: The fire detection and extinguishing system on the Mi-8 helicopter is tasked with detecting, alerting, and extinguishing fires when they occur. The extinguishing process is carried out by discharging extinguishing bottles to the localized fire zones automatically, or manually by pressing the control buttons. To improve the quality of education and training, strengthen theoretical foundations, and enhance the practical operating and troubleshooting skills of trainees, the authors utilized hardware design and microcontroller applications to successfully design and fabricate a training simulator for the fire detection and extinguishing system on the Mi-8 helicopter.

DOI: https://doi.org/10.5281/zenodo.21131289

Traction Motor Trends For Indian Passenger EVs

Authors: Jiteshkumar Dalwala, Dr.R.T Ugale

Abstract: The rapid electrification of India's four-wheeler (4W) passenger vehicle market presents unique powertrain engineering challenges, characterized by stringent cost constraints, dense urban driving conditions, and a strategic imperative to minimize reliance on imported rare-earth materials. This paper presents a comprehensive analysis of traction motor trends within the Indian electric vehicle (EV) sector, focusing on the strategic selection rationale and the resulting drivability dynamics. A critical comparative study is conducted between the currently dominant Interior Permanent Magnet Synchronous Motors (IPMSM) and the highly relevant alternative, AC Induction Motors (ACIM). Furthermore, the paper highlights how advanced dynamic control strategies specifically Field-Oriented Control (FOC) are implemented in high-power simulation environments to mitigate the inherent drivability limitations of each architecture. The findings conclude that while IPMSMs offer unmatched volumetric efficiency and low-end torque for urban commuting, ACIMs provide a vital, supply-chain-secure alternative with superior high-speed coasting dynamics. Ultimately, optimizing Indian EV powertrains requires balancing localized manufacturing capabilities with sophisticated, closed-loop motor control algorithms.

AI-Powered UI Component Generation: A Natural-Language-to-Code Approach Using Large Language Models

Authors: Vivek Sharma, Dr. Jasbir Kaur, Assistant Professor Sandhya Thakkar

Abstract: Front-end developers spend a significant portion of their time on repetitive UI markup, hindering productivity and innovation. This paper presents an AI-powered system that converts plain-English descriptions into styled HTML components using Google’s Gemini language model. The tool supports three styling paradigms: plain CSS, utility-first (Tailwind), and component-based (Bootstrap). It provides an interactive editing environment with a Monaco-based code editor and a live preview pane. In a user study with 50 front-end practitioners, the system reduced average component development time by 81.5% (from 12.4 to 2.3 minutes) and received a System Usability Scale score of 86/100. The paper details the architecture, prompt engineering strategy, implementation, and evaluation. Despite its benefits, the system exhibits limitations, including inconsistent accessibility support, sensitivity to prompt phrasing, and lack of conversational refinement. These findings underline the potential and the remaining challenges of applying generative AI to front-end development.

DOI: https://doi.org/10.5281/zenodo.21155938

Power Factor Correction Using Statcom With LCL Filter And PI Controller

Authors: Dighe Akshay Adinath, Prof.V. R. Aranke, Prof. S.S. Hadpe, Prof.S.S. Khule

Abstract: This paper presents the complete design, mathematical modelling, simulation implementation, and rigorous performance analysis of a Static Synchronous Compensator (STATCOM) for dynamic power factor correction in a three-phase 415 V, 50 Hz AC distribution system. A fully designed LCL output filter (L1=5 mH, C=10 µF, L2=2 mH, Rd=4 Ω, f_res=1.3 kHz), cascaded dual-loop dq-frame PI control architecture, and Synchronous Reference Frame Phase-Locked Loop (SRF-PLL) are incorporated and validated using MATLAB/Simulink R2021b with Simscape Electrical. Simulation results demonstrate: reactive power compensation from Q_Grid=100 kVAR to 0–5 kVAR (>95% efficiency); power factor improvement from 0.894 lagging to unity (1.0); load-side current THD of 0.01% and load voltage THD of 0.02%. P_Stat ≈ 0 kW confirming purelyreactive operation; DC bus stable at ~700 V; and dynamic step-change response within 2–3 AC cycles (60–80 ms).

DOI: http://doi.org/10.5281/zenodo.21157124

mg src=”https://www.ijsret.com/images/download.png” />

Design and Control of a 3-DOF Robotic Arm Using Potentiometer Feedback

Authors: Ankita kumari

Abstract: This research paper presents the design, development, and implementation of a 3-Degree of Freedom (3-DOF) robotic arm using potentiometer-based feedback and an Arduino controller. The system focuses on achieving accurate position control through closed-loop feedback. The proposed model integrates mechanical design, kinematic modeling, electronic interfacing, and embedded programming. Potentiometers are used to monitor joint angles and improve precision and repeatability. The system is cost-effective and suitable for educational and light industrial applications.

Sinusoidal Support Surface Translation versus Cawthorne-Cooksey Exercises for Unilateral Vestibular Hypofunction: A Quasi-Experimental Study

Authors: Professor B. Gopika Ramya

Abstract: Background- Unilateral Vestibular Hypofunction (UVH) results in significant balance deficits, gait unsteadiness, and a high perception of subjective disability. While traditional Cawthorne-Cooksey exercises are a clinical mainstay relying on habituation, mechanical interventions like sinusoidal support surface translations (SSST) offer a technology-driven, specialized approach to challenging the vestibulo-spinal reflex (VSR) and facilitating sensory re-weighting. Objective- To compare the efficacy of traditional Cawthorne-Cooksey exercises and Sinusoidal Support Surface Translation exercises in patients presenting with unilateral vestibular deficits using the Dizziness Handicap Inventory (DHI) and Berg Balance Scale (BBS). Methods- A prospective, quasi-experimental study was conducted at the JKKMMRF College of Physiotherapy, Komarapalayam, with 20 participants (n = 10 per group) selected via purposive sampling. Group A performed the Cawthorne-Cooksey protocol involving progressive head, eye, and body movements. Group B underwent instrumental rehabilitation via mechanical sinusoidal support surface translations focusing on postural stability and rhythmic perturbations. Both protocols spanned 6 weeks (3 sessions/week). Baseline and post-intervention outcomes were evaluated using the DHI and BBS, and data were analyzed via paired and unpaired t-tests (p < 0.05). Results- Both groups showed significant within-group improvements (p < 0.001) after 6 weeks. However, Group B demonstrated significantly greater post-intervention gains than Group A. Group B achieved a superior post-intervention mean BBS score of 41.0 pm 1.89 compared to 32.4 pm 3.95 in Group A (p < 0.0001). Subjective handicap on the DHI also reduced more substantially in Group B (34.4 pm 6.67) than in Group A (48.2 pm 4.81). Conclusion- Sinusoidal support surface translation is a superior clinical intervention compared to Cawthorne-Cooksey exercises, offering more effective outcomes in restoring objective postural stability and mitigating the subjective impact of dizziness in patients with UVH.

The Indian Mega Food Park Scheme: Objectives, Recent Trends And Prospects

Authors: Lakshmi Kunhikrishnan, Elanchezhian J.

Abstract: The demand for food and value-added healthy products in the world grows at a vast rate every year. However, the post-harvest loss (PHL) in the Indian agricultural sector is a great dip to the value addition and the country’s economy. The Government of India (GOI) thus proposed a vision known as Mega Food Park Scheme (MFPS) to couple farmers to the food processors. This assists in reducing the PHL, thereby improving the manufacturing infrastructure, direct and indirect employment, export rate, etc. The main purpose of this review is to illustrate the scheme of the MoFPI and disclose the working milieu of the mega food parks based on the primary objectives set by the Mega Food Park Scheme, their growth to the agricultural sector and the generation of value addition to the consumers.

DOI: http://doi.org/10.5281/zenodo.21206680

A Review of Hybrid Deep Learning Architectures for Real-Time Network Traffic Anomaly Detection: From Signature-Based Systems to Cross-Stream Attention CNN-BiLSTM Frameworks

Authors: Rohan Singh, Assistant Professor Jitender Kumar

Abstract: The proliferation of encrypted network traffic and the accelerating sophistication of cyber-attack methodologies have rendered signature-based intrusion detection fundamentally inadequate for contemporary Security Operations Centre (SOC) and Network Operations Centre (NOC) environments, a problem underscored by a global cyberattack cost exceeding eight trillion United States dollars in 2023. This paper reviews the methodological evolution of network traffic anomaly detection, tracing the progression from signature-matching platforms through statistical and behavioural baselining to deep learning architectures spanning convolutional neural networks, recurrent and attention-augmented sequence models, graph neural networks, and Transformer-based classifiers. Particular attention is given to the limitations of single-modality architectures, which capture either the spatial packet-level structure or the temporal session-level dynamics of network traffic but not both simultaneously, and to the comparative strengths and weaknesses of early, late, and attention-based multi-stream fusion paradigms. The review further examines loss-level strategies for the severe class imbalance characteristic of intrusion-detection benchmarks, and surveys the explainability mechanisms required for analyst trust and MITRE ATT&CK alignment in production SOC deployment. Based on the gaps identified, the paper discusses an emerging direction—lightweight cross-stream attention CNN-BiLSTM frameworks exemplified by TrafficGuardNet—and outlines future research priorities, including model compression for ultra-high-throughput deployment, continual learning under concept drift, and feature-engineering adaptations for TLS 1.3-encrypted environments.

DOI: https://doi.org/10.5281/zenodo.21188906

Formulation and Evaluation of Herbal Nail Lacquer

Authors: Assistance Professor Ms. Smita Mane, Mr. Saad Yusuf Shaikh, Ms. Renuka Ananda Waghmode, Dr. Vijaykumar Kale, Dr. Mahesh Thakare

Abstract: Paronychia is a highly prevalent, painful inflammatory infection involving the lateral and proximal nail folds of fingers and toes, frequently caused by bacterial pathogens such as Staphylococcus aureus or opportunistic fungi such as Candida albicans.17 Conventional treatment strategies predominantly include systemic antibiotics, oral antifungals, or topical creams.6 However, oral medications are often associated with systemic side effects, drug-drug interactions, and hepatic stress, while conventional topical creams are easily washed or rubbed off, leading to poor localized drug bioavailability through the highly keratinized nail plate.6 This study focuses on the design, formulation, and evaluation of a therapeutic, herbal nail lacquer to establish a sustained-release, non-invasive transungual drug delivery system for paronychia.3 An herbal active blend containing clove extract, neem extract, tea tree oil, and aloe vera gel was selected to provide synergistic antibacterial, antifungal, anti-inflammatory, and wound-healing properties directly at the infection site.7 The lacquer base was optimized utilizing nitrocellulose as the primary film-former, ethyl acetate and butyl acetate as the volatile solvent system, dibutyl phthalate and camphor as plasticizers, and isopropyl alcohol as a co-solvent.11 Four trial batches (F1 to F4) were developed with varying concentrations of polymers and active herbal extracts.5 The developed formulations were systematically evaluated for key physicochemical parameters, including drying time, viscosity, non-volatile content, water resistance, film smoothness, gloss, pH, and stability, alongside in vitro antimicrobial efficacy against Staphylococcus aureus and Candida albicans.5 The optimized formulation, batch F3, formed a smooth, glossy, and uniform film with an acceptable drying time of 75 seconds pH of 6.3 and a non-volatile content of 21.5%.14 F3 demonstrated excellent water resistance and a viscosity of 142 cp , ensuring ideal brushability and adherence.14 Furthermore, microbiological assays of F3 revealed a substantial zone of inhibition against both S. aureus (22mm) and C. albicans (24mm), which was highly comparable to commercial synthetic formulations.5 Accelerated stability studies confirmed the integrity of the formulation over 30 days under standard ICH storage conditions.3 These findings indicate that the developed polyherbal nail lacquer is a promising, safe, and cosmetically acceptable transungual delivery platform for managing paronychia.24

The Framework of Fast-Math: Vedic Sutras as the Superior Pedagogical Tool

Authors: Dr. Sangita B. Pimpare

Abstract: This research explores how Vedic Mathematics—a system rediscovered by Swami Bharati Krishna Tirthaji from 1911 to 1918—stacks up against modern arithmetic. Focusing on core sutras such as Urdhva Tiryagbhyam and Nikhilam Navatashcharamam Dashatah, the study compares their computational speed and hardware demands to those of standard long-form methods. The results show that Vedic algorithms reduce the number of intermediate steps and partial products, particularly benefiting high-speed digital systems; implementations in VLSI and FPGA environments, for instance, exhibit reduced critical-path delays and more efficient logic utilization (Yadav et al., 2024; Patel et al., 2015). The paper also critically assesses the teaching effectiveness of Vedic Mathematics—built around sixteen ancient Sanskrit sutras—relative to the conventional arithmetic methods used in schools worldwide. While typical math instruction emphasizes memorization and rigid carry-and-borrow procedures, the Vedic system takes a multidimensional, pattern-recognition approach that more naturally suits how people think. Through a side-by-side comparison of computational speed, error rates, and cognitive load—pitting standard methods against Vedic techniques like Urdhva-Tiryagbhyam (crosswise multiplication) and Nikhilam (base-based subtraction)—the study finds that the Vedic framework cuts intermediate steps by up to 60%, greatly reducing the mental strain of tackling multi-digit problems.

Application of the Discrete Vortex Method to the Calculation of Aerodynamic Characteristics of Coaxial Helicopter Rotors

Authors: Tat Dat Dong, Van Huong Ngo, Phan Trong Sơn, Phan Lê

Abstract: This paper presents the application of the discrete vortex method to calculating the aerodynamic characteristics of coaxial helicopter rotors. The study focuses on modeling the aerodynamic interaction between the upper and lower rotors, which rotate in opposite directions and create a complex unsteady flow field. Based on the discrete vortex method, the rotor blades are replaced by bound and free vortex segments, allowing the determination of pressure distribution, aerodynamic forces, moments, and thrust coefficients. A computational program was developed to evaluate the influence of initial azimuth angle and horizontal hinge distance on the aerodynamic performance of coaxial rotors. The obtained results provide a scientific basis for selecting geometric and kinematic parameters, supporting aerodynamic analysis, rotor design, and digital transformation in aviation education and research.

DOI: https://doi.org/10.5281/zenodo.21235902

Zero-Threshold Ergonomics: Forensic Detailing of Flush Transitions for Fall Prevention in Indian Senior Housing

Authors: Mihir A. Nagpure Has Published A Paper in Guidance of, Prof. Ar. Malini Nathe , and Prof. Ar. Gulfam Shaikh

Abstract: Geriatric slips, trips, and falls constitute the leading cause of accidental trauma and mortality among elderly populations globally, with heightened vulnerability observed within domestic residential environments. While the National Building Code (NBC 2016) of India outlines macro-level barrier-free guidelines, conventional construction practices frequently default to step-downs, raised door thresholds, and deep shower curbs to manage hydraulic drainage. This paper provides a forensic architectural analysis of zero-threshold (flush) ergonomic detailing as a primary architectural intervention for fall prevention. By evaluating the structural, material, and hydraulic intersection of flush-floor transitions, sliding track integrations, and concealed linear trench drainage systems, this study establishes a scalable technical framework. The findings demonstrate that removing micro-spatial steps directly optimizes wheelchair, walker, and assisted-locomotion kinematics while maintaining rigorous waterproof integrity in high-moisture tropical contexts.

DOI: https://doi.org/10.5281/zenodo.21255222

Creating a Sense of Place in Contemporary City Centres: A Framework for Integrating Cultural Identity into Modern Urban Design

Authors: Nensenlo Kath has published a paper in guidance of , Dr. Sudhir V. Dhomane

Abstract: City centres are the social and cultural core of urban areas, but rapid urbanization and globalization have caused many to lose their unique identity. Standardized planning and architecture often result in spaces that lack local character and community connection. Creating a strong sense of place has therefore become an important goal in contemporary urban design. This paper examines how cultural identity can be integrated into modern city centre design to create distinctive and people-oriented urban spaces. It discusses the concept of sense of place, the key elements that shape place identity, and selected case studies. Based on these findings, the study proposes a design framework that balances modernization with local culture. The research concludes that integrating cultural identity into contemporary city centres can strengthen community belonging, improve public spaces, and create more memorable urban environments.

DOI: https://doi.org/10.5281/zenodo.21255338

Architecture as an Experience: Understanding Human Engagement Through Space

Authors: Thejano Pearl Kikon has published a paper in guidance of, Dr. Sudhir V. Dhomane and , Professor Anand A. Pande, Prof. Anand A. Pande

Abstract: Architecture is more than the creation of functional and visually appealing buildings; it shapes how people perceive, experience, and connect with their surroundings. This study explores architecture as a lived experience by examining how perception, memory, emotion, narrative, symbolism, and atmosphere influence human interaction with space. Drawing on phenomenological theories and selected architectural case studies, the research investigates how these experiential dimensions contribute to meaningful architectural environments. A qualitative research approach based on literature review and case study analysis is used to identify principles that support human-centered design. The study argues that architecture gains significance through lived experience rather than physical form alone. It concludes by proposing design principles that help create spaces that are emotionally engaging, culturally meaningful, and responsive to human needs.

DOI: https://doi.org/10.5281/zenodo.21255596

Governance-Aware Agentic AI for Enterprise Engineering Systems: A Design-Science Reference Architecture and Quantitative Risk-Control Model

Authors: Kwan Hong Tan

Abstract: Agentic artificial intelligence is moving enterprise engineering from passive decision support to autonomous task execution, tool use, workflow coordination and continuous optimization. This transition creates a design problem that is neither purely technical nor purely managerial: enterprises need AI agents that improve productivity while remaining auditable, bounded, reversible and accountable. Existing AI governance standards provide essential principles and process requirements, yet many organizations still lack an implementable system architecture that translates these requirements into operational controls at the level of agent planning, tool invocation, data access, human escalation and post-deployment monitoring. This paper develops a governance-aware reference architecture for agentic AI in enterprise engineering systems and formalizes it through a quantitative risk-control model. Using a design-science methodology, the study synthesizes AI risk management standards, algorithmic auditing literature, human-centered AI design and recent work on AI-form organizations, distributed responsibility and AI stakeholder recognition. The resulting artifact, termed the Governance-Aware Agentic AI Control Architecture, integrates six layers: strategic intent and risk appetite, agent orchestration, data-model-tool infrastructure, governance control plane, observability and evidence, and human escalation with adaptive assurance. The paper introduces three formal constructs – Productivity-Adjusted Residual Risk, Governance Debt and Human Override Threshold – to guide deployment decisions. An illustrative scenario evaluation across customer service, finance operations, human-resource screening and supply planning demonstrates how the model can convert abstract governance principles into measurable engineering checks. The contribution is a practical and theoretically grounded architecture for organizations seeking to deploy agentic AI responsibly without reducing governance to after-the-fact compliance documentation. The paper concludes that trustworthy enterprise AI requires control systems designed into the architecture itself, not merely ethical statements attached to autonomous workflows.

Multi-Dimensional Cyber Threat Profiling Using System and Web Logs for Regional, Behavioural, and Risk-Based Attack Analysis

Authors: Ms. Tejaswini Vikas Patil, Professor Dr. R. N. Patil

Abstract: Cybersecurity monitoring is essential for protecting digital systems from web-based attacks, unauthorized access, and suspicious user activity. Traditional log monitoring usually focuses on individual events and does not provide a complete view of attacker behaviour, geographical origin, risk severity, and response requirements. This paper presents a real-time multi-dimensional cyber threat profiling system that analyses system and web logs for regional, behavioural, and risk-based attack analysis. The developed system uses Logstash for real-time log ingestion and processing, OpenSearch for storing processed security events, and OpenSearch Dashboards for visualization. The system processes simulated JSON security events, Apache web server access logs, and SSH authentication logs. It detects attack categories such as brute-force login attempts, SQL injection, cross-site scripting, directory traversal, web scanning, restricted resource access, insider suspicious access, and normal user activity. GeoIP enrichment is used for regional analysis, while a rule-based threat scoring model classifies each event as Low Risk, Medium Risk, or High Risk. High-risk events are stored in a dedicated alert index, and severe events generate response recommendations such as temporary IP blocking with pending analyst review. The implementation demonstrates real-time log ingestion, attack classification, threat scoring, risk classification, alert generation, and dashboard-based threat visualization.

Polyherbal Mouth Dissolving Tablets for Oral Ulcer Management

Authors: Mr.Kundan Kale, Mr.Santosh Waghmare, Ms. Arti Chandanshive

Abstract: Oral ulcers are painful lesions of the oral mucosa that interfere with eating, speaking, and swallowing, significantly affecting quality of life. Conventional therapies, including corticosteroids, antiseptic mouthwashes, and analgesics, often provide only temporary relief and require frequent application, with the possibility of adverse effects. To overcome these limitations, the present study focuses on the development of a polyherbal mouth dissolving tablet (MDT) containing Glycyrrhiza glabra, Curcuma longa, Azadirachta indica, Aloe vera, and clove oil, selected for their anti-inflammatory, antimicrobial, antioxidant, analgesic, and wound-healing activities. The herbal extracts were subjected to phytochemical and compatibility studies before formulation. MDTs were prepared by the direct compression technique using suitable excipients and evaluated for pre-compression properties, post-compression characteristics, in vitro drug release, and antimicrobial activity against selected oral pathogens.

DOI: https://doi.org/10.5281/zenodo.21261380

Intelligent Bronchiectasis Risk Prediction Using Explainable Machine Learning Models

Authors: Shaik Rahena, Associate Professor Mrs.M.Radhika

Abstract: There is a substantial clinical and health care burden associated with bronchiectasis, a chronic respiratory disorder characterised by frequent exacerbations, persistent symptoms, and frequent hospitalisations. Conventional treatment techniques rely on clinician opinion and assessment based on guidelines, which may not adequately reflect the unique risk heterogeneity of each patient. This research presents a paradigm for bronchiectasis risk classification and intervention assistance using machine learning and routinely gathered clinical data. The purpose of analysing a structured dataset that includes demographic, clinical, etiological, lifestyle, and treatment-related variables is to divide patients into groups with varying degrees of risk. Stratified cross validation testing is used for a variety of machine learning models, including feedforward neural networks, Random Forest, XGBoost, and Logistic Regression. Based on the data, it seems that the baseline clinical characteristics are generally linearly separable, as Logistic Regression provides the greatest prediction performance. Ensemble and neural models work well together and compete well in terms of interpretability and decision-support. Applying explainable AI with SHAP helps to make sense of model predictions and isolate key risk variables. Further patient variables uncovered by unsupervised clustering included exacerbation load and disease duration. The suggested approach combines phenotyping, predictability, and explainability to help with individual-level respiratory health management and data utilisation for treatment decision-making in bronchiectasis.

DOI: https://doi.org/10.5281/zenodo.21261667

Deep Learning Framework for Citrus Leaf Disease Detection

Authors: Battu Swarupa, Assistant Professor M.Prasanna Kumar

Abstract: Many foliar diseases may reduce the quality of yields and overall production, and citrus crops are especially susceptible to these diseases. Conventional manual diagnosis may be tedious, time-consuming, and subjective, especially when symptoms seem visually similar. By developing a deep learning system for the automatic detection and categorisation of diseases affecting citrus leaves using EfficientNet-B7, this study addresses a need in the current market. The model enhances generalisation across multiple disease categories, including canker, black spots, greening, melanose, and healthy leaves, by applying transfer learning, large data augmentation, and fine-tuning. To make it seem like there are real-world differences in lighting, orientation, and background conditions, the dataset is preprocessed using normalisation and augmentation. The model is improved by using AdamW and cosine annealing after it has been trained using weighted categorical cross-entropy. When looking at F1-scores, recall, precision, and accuracy within classes, the experimental findings show that it works very well. On average, 97.2% accuracy is attained across all disease categories. The effectiveness and scalability of the proposed technique are confirmed by comparing it to existing designs. Based on the findings, EfficientNet-B7 may be used to smart precision farming and agricultural diagnostics in order to make better use of available resources.

DOI: https://doi.org/10.5281/zenodo.21261914

Deep Learning-Based Tuberculosis Detection from Chest X-Ray Images

Authors: Mulsa Venkata Sravya, Associate Professor Mrs.M.Radhika

Abstract: Due to a lack of technical improvements in diagnosis and treatment, tuberculosis (TB) continues to impact millions of people worldwide, despite the availability of highly effective medicines. Accurate identification and early diagnosis are crucial for reducing the spread and enhancing treatment results. Traditional diagnostic methods, such as sputum microscopy and culture, are labour-intensive and susceptible to human error since they are carried out by laboratory professionals. The field of deep learning (DL) has lately seen remarkable progress, and it shows great promise for improving and automating the accuracy of diagnoses. A DL-based approach to TB detection in chest X-rays is proposed in our work. Following training on a large dataset, our model outperforms traditional methods with a loss of 8.19 percent and an impressive accuracy of 97.32 percent. With convolutional neural networks (CNN) as its foundation and augmented with transfer learning (SqueezeNet) and explainable artificial intelligence (AI) methods like Grad-CAM, the model can accurately identify TB-related patterns while producing few false positives. More reliable, scalable, and rapid solutions for healthcare systems throughout the world may be possible with this approach, which would revolutionise TB diagnosis.

DOI: https://doi.org/10.5281/zenodo.21262210

Hybrid Deep Learning Framework for Electrical Energy Consumption Forecasting

Authors: Paruchuri Raghava Rani, Associate Professor Ch.Naveen

Abstract: Institutions that distribute power and companies that produce electricity, whether public or commercial, rely heavily on long-term electricity demand estimates. It helps with strategic decision-making to improve the quality of energy output and ensures optimal energy utilisation. Countries that have endured energy shortages for a long time, like Iraq, have a pressing need for this. Based on daily home power usage statistics collected from the Iraqi Ministry of power for the Rusafa district of Baghdad between 2022 and 2024, this research draws its conclusions. We also added meteorological data from the same years, which includes things like humidity, sun radiation, and temperature, all of which are outside influences on consumption habits. This paper presents a hybrid model for forecasting that combines LSTM and CNN-based deep learning architectures, an improved stacked hybrid model that uses CNN, GRU, Stacked Bi-LSTM, and machine learning regressors like XGBoost and LightGBM, and so on. The goal of training these models is to enhance energy acoustic production techniques and provide more accurate forecasts. Using metrics like mean relative absolute error (MAPE) and mean root mean square error (RMSE), we trained and assessed the proposed model across 30 epochs to assess the accuracy of the predictions. Our hybrid model utilising the LightGBM regressor outperformed all others examined, with a MAPE of 0.185155 and an RMSE of 0.094603 for the next spilt time period of periodic predictions, respectively. The findings highlight the promise of hybrid modelling approaches for improving power distribution system optimisation and energy forecasting.

DOI: https://doi.org/10.5281/zenodo.21273091

A Hybrid Deep Learning Framework for the Detection of Pancreatic Tumours in CT Images

Authors: P V S R Saiharsha, Assistant Professor M.Prasanna Kumar

Abstract: The aggressive nature of pancreatic cancer and the lack of viable early detection technologies make it a significant health danger. By combining deep learning with image processing methods, this study presents a new framework for accurately detecting pancreatic tumours. As part of the suggested system, CT images are enhanced using CLAHE to bring out more features and contrast in the tumor's more nuanced areas. Tumour segmentation using a U-Net design subsequently enables pinpoint localisation of cancerous tissue. To further optimise learning and model convergence, a CNN classification network is used with a Stochastic Gradient Descent with Momentum optimiser. After following the aforementioned procedures, an end-to-end implementation was evaluated on a dataset consisting of 1,000 pancreatic CT images in MATLAB. A convolutional neural network (CNN) optimised using Stochastic Gradient Descent with Momentum, U-Net for segmentation, and CLAHE-based augmentation, all implemented in MATLAB. SGDM CLAHE is used for preprocessing. By improving picture quality, accurately defining tumour borders, and consistently recognising the existence of a tumour, this framework will aim to overcome several constraints of computed tomography imaging. To help radiologist make quick and accurate diagnosis, which improves patients' survival rates, this study will utilise these methodologies to guarantee a strong solution that will lead to early pancreatic tumour identification. Enhanced convolutional neural network (CNN) categorisation offers a quick, automated, and reliable method that might become a valuable CAD tool for the early diagnosis of pancreatic tumours, leading to better clinical outcomes.

DOI: https://doi.org/10.5281/zenodo.21273244

Cotton Weed Classification Using Hybrid Deep Learning Models

Authors: Talluri Milka, Assistant Professor Mrs.M.Sivaparvathi

Abstract: In order to implement efficient weed control tactics, cotton weed categorisation is an important component of precision agriculture. Hybrid deep learning models provide a strong foundation for this classification challenge by combining several models. This method is taken to the next level with the use of active learning algorithms, which streamline the classification process by cutting down on the size of labelled datasets. Particularly for specialised jobs like weed detection in agricultural applications, the time and money spent manually labelling huge datasets may be substantial; active learning attempts to alleviate this problem. It maximises the efficiency of human annotators by picking the most informative samples to label.

DOI: https://doi.org/10.5281/zenodo.21273393

Cloud Computing Security in the Age of AI: A Systematic Review Using Artificial Intelligence as Both Research Instrument and Object of Study

Authors: Uwaisu Abubakar Umar, Ibrahim Haruna Ibrahim, Aliyu Aminu Dahiru, Emmanuel Martin Teman

Abstract: The rapid migration of organizational infrastructure to cloud computing has introduced unprecedented scalability and efficiency alongside increasingly complex security vulnerabilities. This review examines the evolving relationship between artificial intelligence (AI) and cloud security through an AI-Augmented Systematic Literature Review (ASLR), in which AI serves simultaneously as a research instrument for synthesizing literature and as the object of study within cloud security architectures. Drawing on studies published between 2020 and 2026 sourced from IEEE Xplore, ACM Digital Library, and SpringerLink, the review addresses three objectives: examining how cloud infrastructure supports the deployment of resource-intensive machine learning (ML) models, investigating AI's role as an adaptive defensive layer, and identifying new security vulnerabilities introduced by embedding AI into cloud networks. Findings show that cloud platforms enable resource-intensive ML through elastic compute allocation, architectural abstraction, and intelligent orchestration; that AI strengthens defense through predictive threat detection, continuous policy adaptation, and privacy-preserving federated learning; and that this same integration introduces distinct new vulnerability classes, including agent privilege escalation, lifecycle-specific exploitation, cascading cross-layer failures, serverless architectural weaknesses, and adversarial manipulation. The review concludes that AI functions as a double-edged capability within cloud ecosystems, simultaneously strengthening and expanding the attack surface, and argues for lifecycle-aware, zero-trust security models validated under real-world rather than solely simulated conditions.

DOI: https://doi.org/10.5281/zenodo.21274264

Adaptive Resonant Wireless Power Transfer System For Electric Vehicle Charging Under Coil Misalignment Using MATLAB/Simulink

Authors: Mini, Garvit Sharma, Dr Charu

Abstract: The rapid adoption of electric vehicles (EVs) has created a growing demand for efficient, reliable, and user-friendly charging technologies. Conventional plug-in charging systems require physical connectors, which are susceptible to mechanical wear, environmental degradation, and user inconvenience. Wireless Power Transfer (WPT) has emerged as an attractive alternative by enabling contactless energy transfer through magnetic coupling. However, practical implementation of WPT systems is challenged by coil misalignment, variation in air gap, frequency deviation, and reduced power transfer efficiency. This paper proposes an adaptive resonant wireless power transfer system capable of maintaining high charging efficiency under lateral coil misalignment. The proposed system employs a resonant compensation network together with an adaptive frequency control strategy to improve power transfer performance during varying operating conditions. A complete mathematical model of the system is developed and implemented in MATLAB/Simulink for performance evaluation. The proposed approach is analyzed under different coil alignment conditions, operating frequencies, and transmission distances. Simulation results demonstrate that the adaptive control technique significantly improves voltage stability, output power, and overall efficiency compared with conventional resonant charging systems. The proposed system offers an efficient and practical solution for next-generation intelligent electric vehicle charging infrastructure.

DOI: http://doi.org/10.5281/zenodo.21275702

A Comparative Study Of Face Detection Using Haar Cascade And YOLOv8

Authors: Prithviraj Eknath Kharade, Kashish Dara

Abstract: Face detection is a fundamental computer vision task with applications in surveillance, biometric authentication, attendance systems, and human–computer interaction. This paper presents a comparative study of the traditional Haar Cascade algorithm and the deep learning-based YOLOv8 model. Both methods are evaluated based on accuracy, speed, computational efficiency, and performance under different conditions such as illumination, occlusion, and pose variations. The analysis shows that Haar Cascade is suitable for lightweight, real-time applications with limited resources, while YOLOv8 provides higher detection of accuracy and robustness in complex environments. This study highlights the strengths, limitations, and practical applications of both approaches to assist in selecting an appropriate face detection method for different use cases.

Comparative Analysis of Low-Light Image Enhancement Techniques and Their Impact on YOLOv8 Object Detection

Authors: Arjit Sasan, Abhishek Bhardwaj, Dev Karan Singh, Gurdeep Singh Panwar

Abstract: Low-light conditions present significant obstacles in the design of computer vision solutions, considering that dark images have poor visibility, reduced contrast, unseen details, and higher noise levels, which result in reduced object detection system performance. This research proposes Lumnia, a low-light image enhancement and object detection solution that utilizes multiple image enhancement algorithms and YOLOv8 object detection. The system employs four enhancement modes, which include CLAHE (Contrast Limited Adaptive Histogram Equalization), Gamma Correction, simplified Zero-DCE inspired image enhancement technique, and Auto Enhancement mode. The enhanced image is then sent for further processing using YOLOv8 to detect objects. In addition, the solution uses several image quality metrics, such as MSE (Mean Squared Error), PSNR (Peak Signalto-Noise Ratio), and SSIM (Structural Similarity Index) to evaluate the image enhancement process and assess the differences between the original and enhanced images. The project was written using Python programming language, and it makes use of Flask web framework, OpenCV library, NumPy package, scikit-image, and Ultralytics YOLOv8. Experiments reveal that various image enhancement methods result in varied image visibility and object detection results.

DOI: https://doi.org/10.5281/zenodo.21292904

Crop Disease Prediction Using Convolutional Neural Network (CNN) for Uttarakhand Farmers

Authors: Harsh Rana, Swati, Priyanshu Vats, Aakash Kumar

Abstract: Agriculture sustains over 70% of Uttarakhand's hill-dwelling population, yet crop diseases cause estimated annual losses of ₹1,200 crore across the state (Directorate of Agriculture, Uttarakhand, 2022). Timely, accurate identification of plant diseases is critical to reducing these losses, but access to agronomists in remote hill districts such as Pithoragarh, Chamoli, and Uttarkashi remains severely limited. This paper proposes a custom Convolutional Neural Network (CNN) architecture optimised for crop disease detection under the specific agro-climatic, lighting, and device-availability conditions of Uttarakhand. The model was trained on a curated dataset of 15,850 annotated leaf images spanning six major crops — wheat, rice, tomato, potato, apple, and maize — covering 28 distinct disease classes. Images were sourced from the publicly available PlantVillage benchmark as well as original field photographs collected across five Uttarakhand districts in collaboration with local Krishi Vigyan Kendras (KVKs). The proposed architecture incorporates four convolutional blocks with Batch Normalisation, GlobalAveragePooling, and Dropout regularisation, yielding an overall classification accuracy of 96.7%, precision of 96.2%, recall of 95.9%, and F1-score of 96.0% on the held-out test set. These results outperform all evaluated baselines — SVM (78.3%), Random Forest (82.5%), VGG-16 (88.9%), ResNet-50 (91.4%), and MobileNetV2 (93.2%). The trained model was converted to TensorFlow Lite (TFLite) format and integrated into a prototype Hindi-English Android application named KrishiRakshak, which supports fully offline inference on low-end devices in under 1.5 seconds. A pilot field study with 50 farmers in Pauri Garhwal and Almora districts demonstrated an 82% correct disease identification rate using the application, compared to 47% through unaided visual inspection.

DOI: https://doi.org/10.5281/zenodo.21293194

Smartvision Road Safety System Animal Detection Using Machine Learning

Authors: Jahnvi Singh, Ilma Abrar, Prince Kumar

Abstract: Over the years, accidents caused by animals crossing the road unexpectedly have remained a significant cause of road deaths. Roads near forests are often dark and dense, making it hard for drivers to see animals clearly. Truck drivers particularly struggle with blind spots. This paper proposes a model that can effectively detect animals and alert drivers. We use a deep learning algorithm to identify animals using a large open-source dataset. Our model employs convolutional neural networks to predict objects in each image frame taken from a live camera. If the system identifies an object as an animal, it provides a three-second alert to warn the driver about the approaching animal. This model is not limited to a few animals; since the dataset is open-sourced, the range of detected animals keeps expanding. The model achieves 91% accuracy.

DOI: https://doi.org/10.5281/zenodo.21308595

Web-Based Real-Time Chat System

Authors: Lakshyata Varshney, Shubhi, Disha Bora, Subhashita Kumari, Mr. Ashish Kumar

Abstract: Today's digital age relies on real-time communication systems for quick info sharing across distances and groups. Think instant messaging, team projects, customer service, and social media – they all need fast, reliable, and safe real- time setups as tech advances, people want smooth, quick, and easy communication that works on any device. Standard web apps that use the request-response setup just don't cut it for real-time stuff. This paper looks closely at building a real time chat system app using the MERN stack: MongoDB, Express.js, React.js, and Node.js. The app focuses on safe logins, private and group chats, fast message sending, user profiles, and easy-to-use interfaces. We use WebSocket with Socket.IO for quick, two-way data flow between users and servers. This project focuses on building something practical that fits real-world needs. We start with a look at real-time systems, web communication, and full-stack JavaScript. Then, we talk about how we gathered requirements, designed the system, modeled the database, built the front and back ends, handled real-time events, and added security. Tests show the system works well for reliable real-time communication with little delay and good usability. We then talk about how the system did, how it can grow, and what users thought based on what we saw. In short, we prove that the MERN stack is a great choice for real-time web apps. Future steps could include encryption, AI, and using cloud services for larger groups. This paper is a helpful reference for students, researchers, and developers in real-time web app creation.

DOI: https://doi.org/10.5281/zenodo.21306114

Blockchain-Based Secure Electronic Voting System

Authors: Nikhil Sharma, Raj Gupta, Kunal Tomar, Aman

Abstract: The transition from traditional paper-based voting to electronic systems has introduced significant efficiencies but has simultaneously created centralized vulnerabilities, including susceptibility to database manipulation and a lack of transparent audit trails. This research proposes a decentralized, blockchain-based voting framework designed to restore public trust through cryptographic immutability and end-to-end verifiability. By utilizing a Permissioned Proof of Stake (PPoS) consensus mechanism, the system achieves the high transaction throughput necessary for national-scale elections while maintaining a decentralized security posture that prevents any single entity from compromising the results. The technical core of this framework integrates Zero-Knowledge Proofs (ZKPs) to resolve the tension between voter anonymity and auditability. This allows voters to prove their eligibility and the validity of their ballot without disclosing their identity or specific choice, thereby upholding the sanctity of the secret ballot. To address modern security threats, the study incorporates Post-Quantum Cryptography (PQC) to safeguard against future decryption capabilities and utilizes Layer 2 scaling solutions to ensure network resilience during peak voting periods. Methodological validation was conducted through a simulated electoral environment, testing the system against common attack vectors such as DDoS and 51% attacks. The results indicate that the decentralized model significantly reduces the risk of systemic fraud compared to centralized alternatives. This paper concludes that while socio-technical barriers to entry exist, the proposed blockchain architecture provides a scalable, secure, and transparent foundation for the future of digital democracy.

DOI: https://doi.org/10.5281/zenodo.21306064

An End-to-End Deep Learning Framework for Sugarcane Quality Assessment Based on Near-Infrared Spectroscopy and Explicit Feature Interaction-Aware Graph Neural Networks

Authors: Parag Yadav, Assistant Professor Raj Kumar, Shantanu Yadav, Tanya Rana, Yash Singh Kathayat

Abstract: Accurate, rapid, and non-destructive assessment of sugarcane quality is a critical requirement for optimizing harvesting schedules, refinery throughput, and fair payment systems in the global sugar industry. Conventional wet chemistry methods, while accurate, are time-consuming, reagent-intensive, and unsuitable for real-time field deployment. Near-Infrared (NIR) spectroscopy has emerged as a powerful analytical technique for non-destructive quality measurement; however, existing machine learning approaches applied to NIR spectral data predominantly rely on shallow models or architectures that treat samples in isolation, thereby neglecting the rich relational and interaction structure latent within spectral populations. In this paper, we propose a novel end-to-end deep learning framework, designated EFI-GNN-NIR, that integrates 1D Convolutional Neural Network (CNN)-based spectral feature extraction with an Explicit Feature Interaction-Aware Graph Neural Network (EFI-GNN) for simultaneous prediction of Brix percentage, Pol percentage, and purity index of sugarcane samples. The proposed framework constructs a population-level spectral similarity graph wherein each node represents an individual NIR sample and edges encode cosine similarity relationships between learned spectral embeddings. An explicit feature interaction module, employing bilinear cross-network operations within the graph message-passing paradigm, enables the model to capture higher-order cross-wavelength dependencies that conventional models overlook. The framework is evaluated on a benchmark NIR sugarcane dataset augmented with agronomic metadata, demonstrating superior performance with R2=0.978R^2 = 0.978 R2=0.978, RMSE=0.142RMSE = 0.142 RMSE=0.142 Brix, and RPD=6.84RPD = 6.84 RPD=6.84 for Brix prediction, outperforming state-of-the-art baselines including Partial Least Squares Regression (PLS-R), Support Vector Regression (SVR), Random Forest (RF), standard CNN, and vanilla GNN architectures. Multi-task learning across three quality parameters yields consistent improvements over single-task counterparts. This work provides a significant contribution toward precision agriculture, smart sugarcane farming, and real-time quality monitoring pipelines deployable at factory intake points.

DOI: https://doi.org/10.5281/zenodo.21306034

AI-based System for Automated Detection of Vehicle Insurance Expiry Dates

Authors: Priya Malik, Tripti kashyap, Nikita Panwar, Jatin Kamboj

Abstract: In order to maintain road safety and legal compliance, it is essential to keep precise track of when auto insurance expires. In this research, an AI-based system that uses sophisticated image recognition techniques to automatically detect and track the expiration dates of motor insurance is proposed. The system uses machine learning (ML) models and optical character recognition (OCR) to extract pertinent data from digital photos of insurance documents, including the policy number, expiration date, and vehicle details. Providing consumers with real-time warnings for insurance renewals, this system's seamless integration with web-based and mobile platforms is one of its primary features. Through comprehensive testing on a wide range of auto insurance document datasets, we show that the system is highly accurate and efficient at extracting expiration dates from a variety of document types and formats.Our findings demonstrate the potential for this AI-driven solution to support larger efforts in automating compliance and improving user convenience in the insurance domain. The suggested system offers scalability for fleet management systems, making it a versatile tool for both individual and enterprise-level use. It also greatly improves upon traditional manual tracking methods by automating the process, reducing human error, and ensuring timely insurance renewals.

DOI: https://doi.org/10.5281/zenodo.21306006

Towards Sentient Companions: A Strategic Framework for Developing Emotionally Persistent, Morally Aware, and Adaptively Intelligent Non-Player Characters Using Large Language Models and Deep Reinforcement Learning

Authors: Rajan Kumar, Vaibhav Verma, Bishwajit Andia, Sakshi Singh

Abstract: Non-Player Characters (NPCs) in Interactive Digital Environments have traditionally been limited by their artificial script, repetitive loops of actions, and primarily lack in psychological depth. The standards players have for NPCs are increasingly advanced, and the role of creating narrative is becoming a key quality indicator for games, making a level of memory, emotions, moral reasoning and strategic adaptations a requirement at this level. The authors introduce a comprehensive framework for research and development of four aspects of next-generation NPC intelligence: Emotional Memory and Relational Persistence, Moral Fatigue and Ethical Exhaustion, Adaptive Villain Intelligence, and Crowd Behavioural Authenticity. This research builds on the theories and computational paradigms of Large Language Models (LLMs) [4,12], Deep Reinforcement Learning (DRL) [13] and hybrid Behaviour Tree architectures [11] to create the theoretical and engineering underpinning needed to develop NPCs that go beyond the traditional role of a background character. Experimental comparisons conducted by [8] show that DRL-based antagonists can give rise to significantly higher player satisfaction and immersion scores when compared to Finite State Machine (FSM) based ones. In addition, the LLM-based crowd agents had better behavioural understandability scores according to the related crowd simulation studies. The paper ends with a few recommendations for researchers who wish to put forth original scientific work in the fast-changing field of Game Artificial Intelligence.

DOI: https://doi.org/10.5281/zenodo.21305953

Spectrophotometric Determination of Mercury (II) Using 3,5-Dimethoxy-4-Hydroxy Benzaldehyde Isonicotinoyl Hydrazone (DMHBIH)

Authors: S. Vidyasagar Babu, M. Jaya Ramudu

Abstract: Mercury (II) is a potent environmental pollutant with significant health risks. This study presents a spectrophotometric method for the determination of trace amounts of mercury(II) using 3,5-dimethoxy-4-hydroxy benzaldehyde isonicotinoyl hydrazone (DMHBIH) as a complexing agent. The method is based on the formation of a yellowish-brown complex between mercury(II) and DMHBIH in a hydrochloric acid-ammonium hydroxide buffer (pH 6.5). The complex exhibits maximum absorbance at 425 nm, and the method obeys Beer’s law in the concentration range of 0.24–2.88 µg/mL. The molar absorptivity and Sandell’s sensitivity were found to be 2.0 × 104 L mol-1 cm-1 and 0.00362 µg/cm-2, respectively. The method was successfully applied to determine mercury (II) in liver samples, demonstrating its applicability in real-world scenarios.

Cybersecurity knowledge Graph from Malware Attacks Action Reports

Authors: Nitish, Bijender, Suraj, Himanshu

Abstract: After Action Reports ( AARs) provide intensive analysis of cyber incidents. Extraction of materials Cyber- knowledge from these sources will provide security researchers with reliable information, which. It can be used to identify or nd patterns of cyberattacks. In this paper we describe a framework to. extract information from AARs, combine similar organizations and aggregate the extracted information, and. They represent extracts from the Cybersecurity Knowledge Graph ( CKG). We remove entities by We are creating a custom named entity detector called Malware Entity Extractor ( MEE). Then we a neural network to predict how pairs of malware entities are related to each other. When we predicted For entity pairs and the relationships between them, we represent the entity-relation set in CKG. Our next one The step in the process is the consolidation of similar organizations, reforming our CKGs. This mixture helps to represent the intelligent fish Extracts from numerous papers and reports. Fused CKG is known from many AARs, as well. Inter- organizational relationships extracted from separate reports. Because of this mix, a security researcher. can handle questions and retrieve better answers on a hybrid CKG, than an unhybrid knowledge article. Wealso showcase the various logic capabilities that can be leveraged by the security analyst to bring us fused C. K. G

DOI: https://doi.org/10.5281/zenodo.21295337

A Survey On Diabetic Retinopathy Image Classification Features & Techniques

Authors: Nand Kishor, Prof. Akrati Shrivastava, Associate Prof. Dr. Jayshree Boaddh

Abstract: The remarkable success of machine learning has prompted interest in its application to medical imaging diagnosis. Even though state-of-the-art deep learning models have achieved human-level accuracy on the classification of different types of medical data, these models are hardly adopted in clinical workflows, mainly due to their lack of interpretability. This paper has done a deep survey on diabetic retinopathy image classification work done by researchers. Paper has list the features used by the researcher for medical image classification. Further techniques involve in diabetic retinopathy image classification was detailed. Finally various methods of image classification models comparison evaluation parameters were mentioned by the paper.

DOI: http://doi.org/10.5281/zenodo.21305724

Hybrid Architectural Frameworks for Context-Aware Sentiment Analysis in Next-Generation P2P Messaging: Addressing Sarcasm Resolution, Multilingual Slang, and Real-Time Emotional Shift Detection

Authors: Dr. Raj Kumar, Vishal Kumar, Arjun Kumar Shah, Pradeep Kr. Yadav, Priya Saini

Abstract: Modern sentiment classification systems face severe performance degradation when deployed over private peer-to-peer (P2P) communication streams. This drop is primarily driven by the stark divergence between casual conversational language and the structured datasets historically used to train core language models. This study explores the structural and algorithmic requirements needed to sustain high-accuracy affective computing within contemporary messaging networks. We identify three distinct operational failure modes: tokenization failures caused by widespread leetspeak and evolving youth vernacular; polarity flips triggered by contextual emoji usage that traditional text-cleaning steps omit; and the inherent latency-privacy bottleneck brought on by demanding a sub-50-millisecond execution window under end-to-end encryption. To mitigate these vulnerabilities, we introduce the Hybrid Affective Reasoning System (HARS), an architecture that pairs a Parameter-Efficient Fine-Tuned DistilRoBERTa-v2 core with a deterministic Semantic Compositional Lexicon (SCL) and a dependency parser via spaCy for localized Aspect-Based Sentiment Analysis (ABSA). Additionally, we establish Sentiment Velocity (SV) as a dynamic temporal derivative designed for proactive escalation tracking, demonstrating that early-fusion multi-modal attention preserves accuracy against emoji distortion, containing drop rates to under 3%. Our accompanying three-tiered, privacy-first data ingestion framework utilizes HMAC-based tokenization, structural generalization, and differential privacy validated against EDPB Guidelines 1/2026. Testing against a dataset of 100,000 anonymized P2P interactions demonstrates a 91.4% macro-F1 score, yielding an 8.2% improvement over isolated, single-model configurations. Zero-shot evaluations leveraging Pragmatic Metacognitive Prompting (PMP) on large scale models further prove its viability for identifying nuanced linguistic irony.

DOI: https://doi.org/10.5281/zenodo.21307192

Impact of AI Models on Digital Harassment Prevention

Authors: Kapil Dev, Ankit Raj, Apurv Pal, Renu Saini

Abstract: Digital harassment has become one of the most serious challenges in modern online communication systems. The rapid growth of social media platforms, online gaming communities, messaging applications, and digital forums has increased the spread of cyberbullying, hate speech, abusive communication, trolling, identity-based attacks, and online threats. Traditional moderation methods are no longer sufficient because millions of user-generated posts, comments, and messages are uploaded every minute. Artificial Intelligence (AI) has emerged as an effective solution for automated content moderation and harassment prevention. This research paper examines the role of Artificial Intelligence models in detecting and preventing digital harassment. The study analyzes various Machine Learning (ML), Deep Learning (DL), Natural Language Processing (NLP), and Transformer-based models used for toxicity detection and online safety. Models such as Logistic Regression, Naive Bayes, Support Vector Machine (SVM), Long Short-Term Memory (LSTM), Bidirectional Encoder Representations from Transformers (BERT), and GPT-based systems are compared based on accuracy, contextual understanding, scalability, and real-time performance. The paper also discusses AI-powered moderation systems, sentiment analysis, multilingual processing, toxicity detection, behavioral analysis, and ethical concerns related to automated moderation. The study highlights major limitations including algorithmic bias, false positives, privacy concerns, and challenges in understanding sarcasm, humor, and cultural context. The findings show that Transformer-based AI models significantly improve harassment detection accuracy compared to traditional ML approaches. However, responsible AI development and human oversight remain necessary for fair and transparent moderation systems.

DOI: https://doi.org/10.5281/zenodo.21307561

MediBox – AI Powered Medical Diagnosis & Consultation System

Authors: Aman Gupta, Assistant Professor Raj Kumar, Aditya Kumar, Yash Raj, Ankita Singh

Abstract: In recent years, the integration of digital technologies and Artificial Intelligence (AI) in the healthcare sector has significantly transformed the way medical services are delivered. Despite these advancements, many healthcare systems still face challenges such as delayed diagnosis, lack of immediate medical assistance, inefficient handling of medical records, and limited availability of medical professionals, especially in remote and rural areas. These challenges often result in increased health risks for patients and reduced efficiency in medical services. To address these issues, this project titled “MediBox – AI Powered Medical Diagnosis & Consultation System” has been developed. MediBox is a web-based healthcare platform designed to assist patients in obtaining an initial medical diagnosis using AI-based analysis. The system allows patients to upload medical reports, describe their symptoms, and submit health-related information through a user-friendly interface. The AI component of the system processes this data to generate preliminary diagnostic insights, which serve as decision-support information rather than final medical conclusions. A key feature of the proposed system is the integration of human expertise with AI intelligence. All AI-generated diagnostic results are forwarded to registered medical professionals for review and verification. Doctors can analyze the AI suggestions, approve or modify the diagnosis, and provide expert recommendations. This hybrid approach ensures accuracy, reliability, and ethical use of AI in healthcare, reducing the chances of incorrect diagnosis while improving overall efficiency. The MediBox system provides separate dashboards for patients and doctors, ensuring role-based access and secure handling of sensitive medical data. Patients can manage their medical history, view previous diagnoses, and track consultations, while doctors can efficiently manage patient requests and diagnostic reviews. Strong authentication mechanisms and data security measures are implemented to protect patient confidentiality and privacy. In conclusion, MediBox aims to bridge the gap between patients and healthcare professionals by combining Artificial Intelligence with medical expertise. The system enhances accessibility to healthcare services, reduces diagnosis time, and supports better clinical decision-making, making it a reliable and scalable solution for modern healthcare systems.

DOI: https://doi.org/10.5281/zenodo.21307896

Sign Language Recognition Using Deep Learning

Authors: Ms. Sayali Vasantrao Gund, Asst. Prof. S. A. Nandi

Abstract: The rapid growth of urban infrastructure has increased the demand for energy-efficient and sustainable construction materials. This study focuses on the integration of IoT-based temperature sensors within plastic-perlite composite bricks to enhance building energy monitoring and management. The proposed system utilizes waste plastic as a partial replacement material combined with expanded perlite to develop lightweight, thermally efficient, and eco-friendly building units compatible with standard M20-grade concrete structures. IoT temperature sensors (DS18B20) are embedded within the composite bricks to continuously monitor internal and external temperature variations, enabling real-time data collection and analysis through wireless communication modules. Laboratory tests, including compressive strength, thermal conductivity, and water absorption, are conducted to evaluate the structural and thermal behavior of the developed bricks. The experimental results show that the integration of perlite and waste plastic significantly improves thermal resistance, reduces heat transfer, and decreases structural dead load by lowering the average brick density to 821–907 kg/m³ compared to 1800 kg/m³ for conventional units. This research promotes the development of sustainable smart construction materials, aligning with modern green building technologies and the global agenda for energy efficiency.

PawGuardian: An AI-Powered Smart Animal Care and Rescue Ecosystem

Authors: Hritik Sharma, Utkarsh Chauhan, Vishesh Kumar, Sagar Kumar

Abstract: Animal safety and rescue management have become major concerns due to the increasing number of stray animals, lost pets, delayed rescue responses, and limited access to nearby veterinary services. Traditional animal rescue systems often lack real-time communication, centralized monitoring, and intelligent assistance, resulting in inefficient rescue operations and reduced recovery rates. To address these challenges, this research proposes PawGuardian: An AI-Powered Smart Animal Care and Rescue Ecosystem, a modern mobile-based platform designed to enhance animal welfare through intelligent technologies and real-time digital services. The proposed system integrates multiple functionalities including lost pet reporting, nearby animal hospital and shelter detection, emergency rescue assistance, GPS-based location tracking, cloud database management, and instant notification services. The platform utilizes Firebase cloud services, mobile application frameworks, and AI-assisted modules to create a scalable and user-friendly ecosystem for pet owners, rescuers, shelters, and animal welfare organizations. The application enables users to report missing pets, locate nearby rescue resources, receive emergency alerts, and improve communication among rescue communities. Additionally, the proposed system aims to improve rescue coordination efficiency, increase lost pet recovery rates, and promote public participation in animal protection activities. The integration of artificial intelligence and real-time location technologies provides the foundation for future enhancements such as automated pet identification, image-based matching systems, predictive rescue analytics, and smart emergency response mechanisms. This research demonstrates how modern mobile technologies, cloud computing, and AI-driven solutions can be utilized to develop an intelligent, scalable, and socially impactful animal care and rescue platform capable of transforming traditional rescue operations into a smart digital ecosystem.

DOI: https://doi.org/10.5281/zenodo.21308557

Mood Swing Analysis Using Artificial Intelligence

Authors: Ms. Shrutika Suresh More, Ms. Shridevi Amol Nandi

Abstract: Emotion recognition is a pivotal component in the evolution of Human-Computer Interaction (HCI) and digital mental health. Traditional unimodal systems, which rely solely on text or facial expressions, often suffer from low reliability due to the subjective and complex nature of human expression. This paper presents a robust, computationally efficient multimodal framework for Mood Swing Analysis by integrating Natural Language Processing (NLP), Computer Vision, and Spectral Audio Analysis. Our architecture utilizes a multinomial Logistic Regression model for textual sentiment, alongside two distinct 18-layer Deep Residual Convolutional Neural Networks (ResNet18) optimized for facial expressions and Mel-spectrogram-based voice analysis. The unified system is deployed via a Django web framework, offering a centralized interface for triple-modality emotion detection. Experimental validations demonstrate that the late fusion of these distinct modalities mitigates context-ignorance, resolves cross-modal conflicts, and significantly enhances overall classification accuracy on standard hardware footprints.

DOI: http://doi.org/10.5281/zenodo.21308797

A GenAI – Assisted Full Stack Architecture for Enhanced Consumer Engagement

Authors: Mr. Amit Kumar, Aroo Singh, Nishant Koranga, Abhishek Kumar, Aman Raj

Abstract: E-commerce has become a fundamental component of the global digital economy, enabling businesses to reach a wide customer base and provide seamless online shopping experiences. However, traditional e-commerce platforms often face challenges such as lack of personalization, manual content generation, inefficient customer support, and limited user engagement. These limitations can negatively impact customer satisfaction and sales performance. This research presents an intelligent E-commerce platform powered by Generative AI to enhance user experience and automate critical business processes. The proposed system leverages advanced AI models to generate dynamic product descriptions, provide personalized product recommendations, and enable conversational AI-based shopping assistants. Additionally, the system incorporates AI-driven image generation and smart search capabilities to improve product discovery. By integrating Generative AI techniques, the platform aims to reduce manual effort, increase operational efficiency, and deliver a more engaging and customized shopping experience. The proposed approach demonstrates significant potential in transforming traditional e-commerce systems into intelligent, adaptive, and user-centric platforms.

DOI: https://doi.org/10.5281/zenodo.21309173

A Lightweight Deep Learning Framework for Deepfake Image Detection Using Convolutional Neural Networks

Authors: Himanshu, Devkant Singh, Isha Goyal, Satender, Vipin Kumar Dhiman

Abstract: The rapid advancement of generative artificial intelligence has led to the widespread creation of deepfake media, where synthetic images and videos are generated using deep learning algorithms to imitate real individuals. Although such technologies provide significant benefits in entertainment, education, and digital media production, their misuse in misinformation campaigns, identity fraud, cybercrime, and political manipulation has created serious societal concerns. The increasing realism of manipulated content has made manual identification difficult, thereby creating a strong demand for automated and reliable deepfake detection systems. This research presents a lightweight and computationally efficient framework for detecting deepfake facial images using convolutional neural networks and image preprocessing techniques. The proposed system utilizes image normalization, resizing, augmentation, and feature learning through a CNN-based architecture to classify images into real and fake categories. The model is trained on publicly available deepfake datasets and implemented using Python, OpenCV, TensorFlow, and PyTorch-based libraries. Experimental evaluation demonstrates that the proposed approach achieves stable learning behaviour, high classification performance, and reduced overfitting while maintaining low computational complexity. The model effectively identifies visual inconsistencies such as blending artifacts, texture irregularities, and illumination distortions commonly present in manipulated images. The study further discusses challenges associated with deepfake detection, limitations of image-based classifiers, and future opportunities involving transformer architectures, explainable AI, and multimodal detection systems.

DOI: https://doi.org/10.5281/zenodo.21309464

AI-Driven Hospital Readmission Risk Prediction Using XGBoost

Authors: Gaurav Saini, Parkh Kumar, Nikhil Pal, Prashant Kumar, Assistant Professor Rimmy

Abstract: Hospital readmission prediction has emerged as a major research area in the healthcare sector because unplanned patient readmissions significantly increase healthcare expenditure, reduce hospital efficiency, and negatively affect patient outcomes. Readmission generally occurs when a patient is admitted to the hospital again within a short duration after discharge, often due to incomplete recovery, poor follow-up care, or chronic medical conditions. Identifying patients who are at high risk of readmission before discharge is therefore essential for improving treatment quality and optimizing hospital resource management. Traditional healthcare prediction methods are often limited in handling large and complex medical datasets, whereas machine learning techniques provide more efficient and intelligent solutions for predictive healthcare analytics. This research paper presents a machine learning-based hospital readmission prediction framework using the XGBoost classifier. The proposed framework is designed to analyze healthcare datasets and predict whether a patient is likely to be readmitted after discharge. The system includes several important stages such as data collection, data preprocessing, feature engineering, model training, model evaluation, and deployment. During preprocessing, missing values, duplicate records, and inconsistent data are handled to improve data quality. Feature engineering techniques are applied to extract meaningful information from patient records, including demographic details, diagnosis history, laboratory reports, medications, and previous admission records. The XGBoost algorithm is used as the primary classification model because of its high efficiency, scalability, and superior predictive accuracy compared to traditional machine learning algorithms. The trained model is evaluated using various performance metrics such as accuracy, precision, recall, F1-score, and ROC-AUC score to measure its predictive capability and reliability. To provide a practical implementation of the proposed framework, the prediction model is deployed using the Streamlit framework, which allows the development of an interactive and user-friendly web application. Healthcare professionals can use the application to input patient information and instantly obtain readmission risk predictions. This real-time prediction capability can assist doctors and hospital administrators in making informed decisions regarding patient discharge planning, follow-up treatment, and preventive care strategies. Experimental results demonstrate that the proposed machine learning framework achieves strong predictive performance and can effectively identify high-risk patients before discharge. The framework has the potential to reduce avoidable hospital readmissions, lower healthcare costs, improve hospital efficiency, and enhance the overall quality of patient care. Furthermore, this research highlights the growing importance of artificial intelligence and machine learning technologies in modern healthcare systems. Future enhancements may include the integration of deep learning models, real-time electronic health record systems, and cloud-based healthcare analytics for improved scalability and prediction accuracy.

DOI: https://doi.org/10.5281/zenodo.21350968

AI-Driven Personalized Learning Path Recommendation System Using MERN Stack and GenAI

Authors: Shanvi Deep, Priyanshu Roshan, Rishu Raj, Shubham Anand, Mr. Sohan Lal

Abstract: The abundance of online educational resources often overwhelms self-paced learners due to lack of structured guidance. This paper proposes an AI-Driven Personalized Learning Path Recommendation System that generates a complete week-by-week learning roadmap from a single userentered topic. The system automatically decomposes topics into subtopics, arranges them logically, estimates study duration, recommends high-quality resources (videos, documentation, tutorials), and supports realtime progress tracking. Built using the MERN stack (MongoDB, Express.js, React.js, Node.js) and integrated with GroqAI for fast, intelligent generation, the platform delivers personalized, actionable learning plans. Evaluation through user studies and system testing demonstrates high relevance, logical coherence, clarity, and practical usefulness. The system significantly reduces planning effort and improves learning structure for students and self-learners.

DOI: https://doi.org/10.5281/zenodo.21352186

Explainable AI for Life-Critical Healthcare Systems

Authors: Vipul Kumar, Aashu Saini

Abstract: Artificial intelligence (AI) has emerged as a transformative technology in modern healthcare, demonstrating exceptional performance in diagnostics, prognosis, treatment planning, and patient monitoring. However, the deployment of AI in life-critical healthcare systems raises profound concerns regarding transparency, accountability, and trustworthiness. Black-box AI models, despite their high predictive accuracy, fail to provide clinicians with comprehensible reasoning, which is unacceptable in high-stakes medical environments. Explainable Artificial Intelligence (XAI) addresses this critical gap by making AI decision-making processes interpretable and transparent to human experts. This paper presents a comprehensive review of XAI methods applicable to life-critical healthcare systems, examining state-of-the-art techniques such as LIME, SHAP, Grad-CAM, attention mechanisms, and rule-based explanations. We discuss the unique challenges of applying XAI in critical care settings, including real-time explanation requirements, regulatory compliance, clinician trust, and ethical considerations. Furthermore, we explore current XAI applications in medical imaging, clinical decision support systems, ICU monitoring, and drug discovery. Our analysis reveals that while XAI holds immense promise, significant research gaps remain in achieving faithful, robust, and clinically meaningful explanations. We outline future research directions and propose a framework for evaluating XAI systems in healthcare environments.

DOI: https://doi.org/10.5281/zenodo.21353126

AI-Assisted Learning Management System using Retrieval-Augmented Generation and MERN Architecture

Authors: Mohd Abdulla, Navanish Mehta, Abdulla Ansari, Saif Ali Abhishek Tyagi

Abstract: Artificial Intelligence technologies are increasingly being adopted in educational platforms to improve learning accessibility and interactive learning support. Traditional Learn-ing Management Systems (LMS) primarily focus on content delivery and administrative management but often lack adaptive learning capabilities and contextual educational assistance. This paper presents the design and development of a scalable AI-driven Learning Management System that integrates Retrieval-Augmented Generation (RAG), LangChain, and Vector Database technologies to enhance personalized education and student engagement. The proposed system allows users to upload educational documents in PDF format and interact with the content through AI-powered features such as contextual question answering, automatic flashcard generation, intelligent quiz creation, and personalized study assistance. The system utilizes MERN stack architecture for scalable full-stack development, JWT and Google OAuth for secure authentication, MongoDB Atlas Vector Search for semantic retrieval, and LangChain for orchestrating AI workflows. A Flutter-based mobile application is developed using clean architecture and Provider state management to ensure modu-larity, scalability, and maintainability. The implemented RAG pipeline improves contextual accuracy by retrieving relevant document embeddings before generating responses from the Large Language Model (LLM). Experimental analysis demon-strates improved learning accessibility, efficient revision support, and enhanced user interaction compared to traditional LMS platforms. The developed system demonstrates the practical use of AI-assisted educational platforms for improving student interaction, revision efficiency, and contextual learning support.

DOI: https://doi.org/10.5281/zenodo.21354686

AI-Based Early Warning System for Hospital-Acquired Infections

Authors: Sujal Sahu, Mukul Sharma, Gulshan, Vansh Verma, Prabhat Mishra

Abstract: AI-based early warning systems represent a transformative approach to improving hospital infection control by enabling the early detection and prediction of hospital-acquired infections (HAIs) among admitted patients. These systems leverage the computational power of artificial intelligence and machine learning models to analyze structured clinical data, including patient demographics, ICU admission status, duration of hospitalization, use of invasive devices such as ventilators and catheters, and underlying comorbidities. By identifying complex patterns and risk factors in real time, the system provides healthcare professionals with timely risk assessments that support proactive clinical decision-making and targeted intervention strategies. This research presents the design, implementation, and evaluation of an AI-based early warning system specifically developed for predicting HAIs within hospital environments. The study utilizes a dataset of patient records and applies machine learning models such as Gradient Boosting, Logistic Regression, and Random Forest to classify infection risk levels. The system is integrated with a web-based interface to facilitate ease of use for healthcare practitioners. Furthermore, the research examines key challenges including data quality, model interpretability, clinical reliability, and ethical considerations related to patient data privacy. The findings highlight the effectiveness of AI-driven approaches in enhancing patient safety, reducing infection rates, and optimizing hospital resource utilization, thereby contributing to more efficient and proactive healthcare management systems.

DOI: https://doi.org/10.5281/zenodo.21356276

AI-Based Face Recognition Attendance System

Authors: Aditya, Akshat Saxena, Ankit Kumar, Rahul Singh, Shivam Singh Chandel

Abstract: Attendance management plays a significant role in educational institutions for monitoring student participation, maintaining academic records, and ensuring classroom discipline. Conventional attendance systems based on manual entry methods are time-consuming, inefficient, and prone to human errors and proxy attendance. With the rapid advancement of Artificial Intelligence (AI), Machine Learning (ML), and Computer Vision technologies, automated attendance systems using facial recognition have become an effective alternative to traditional approaches. This research presents an AI Based Face Recognition Attendance System designed to automate attendance recording through real-time facial recognition techniques. The proposed system integrates Artificial Intelligence with web-based technologies to identify and verify students using webcam input. The system is developed using React.js for frontend development, Flask APIs for face recognition operations, Node.js and Express.js for backend services, and MongoDB for database management. OpenCV and the Local Binary Pattern Histogram (LBPH) algorithm are used for face detection and face recognition processes. The system provides functionalities such as student enrollment, face dataset generation, automated attendance marking, period-wise attendance management, administrator authentication, attendance dashboards, and attendance log monitoring. Facial images captured through webcam devices are processed using OpenCV algorithms and matched against trained facial datasets to recognize students in real time. Once recognition is successful, attendance records are automatically stored in MongoDB collections and displayed through interactive dashboard interfaces. The proposed framework improves attendance accuracy, minimizes manual workload, reduces proxy attendance, and enhances data management efficiency. Experimental analysis demonstrates that the system performs effectively under moderate lighting conditions and provides reliable real-time recognition performance. The developed solution is scalable, user-friendly, cost-effective, and suitable for smart classrooms, universities, and modern educational environments.

DOI: https://doi.org/10.5281/zenodo.21357214

AI-Powered Chatbots for Healthcare Assistance

Authors: Ayush Kumar, Soumya Rawat, Sunny Kumar, Tushar Tyagi

Abstract: Artificial Intelligence (AI) has significantly transformed the healthcare sector by improving accessibility, efficiency, and quality of services. Among various AI applications, chatbots have emerged as powerful tools for providing healthcare assistance. These AI-powered conversational agents use Natural Language Processing (NLP) and Machine Learning (ML) techniques to interact with users, provide medical guidance, and support healthcare professionals. This research highlights how chatbot systems can reduce the burden on healthcare infrastructure while ensuring timely assistance. Recent systematic reviews indicate that digital healthcare leads AI adoption at 78%, with virtual health assistants and chatbots cited as the top return on investment (ROI) use case for 37% of digital healthcare providers (Benemann, 2026).

DOI: https://doi.org/10.5281/zenodo.21357564

AI Powered Patient Journey Optimization Platform

Authors: Anmol Tyagi, Abhishek Tyagi, Vansh Sharma, Anas Malik, Shivam Tyagi, Dr. Iqbal

Abstract: The rapid advancement of Artificial Intelligence (AI) and cloud-based healthcare technologies has transformed the way medical organizations manage patient communication and service delivery. This research presents an AI-powered healthcare contact center system designed to improve patient engagement, real-time monitoring, and dynamic call prioritization through intelligent computational techniques. The proposed system integrates cloud-native architecture, machine learning models, sentiment analysis, and healthcare interoperability standards to create a smart and responsive communication framework for modern healthcare environments. The system focuses on real-time patient journey mapping, allowing healthcare providers to monitor patient interactions, treatment stages, and emergency conditions more efficiently. By utilizing AI-driven prioritization mechanisms, the framework automatically classifies patient requests based on urgency, emotional state, and clinical context, ensuring that critical cases receive immediate attention. The proposed model also incorporates speech and sentiment analysis to detect stress, discomfort, or emotional instability during patient conversations, thereby enabling proactive medical intervention.To ensure scalability and flexibility, the architecture follows a distributed microservices approach integrated with standardized healthcare data exchange protocols. The framework supports seamless communication between healthcare databases, patient management systems, and intelligent analytical modules. Experimental analysis demonstrates improvements in operational efficiency, response time, patient satisfaction, and healthcare resource utilization when compared with conventional healthcare communication systems. The research highlights the significance of intelligent healthcare communication platforms in reducing delays, improving care continuity, and supporting personalized patient experiences. Furthermore, the study discusses implementation challenges related to data privacy, ethical AI usage, interoperability, and regulatory compliance. The proposed framework offers a future-ready solution for smart healthcare ecosystems and demonstrates the potential of AI-driven communication systems in enhancing the quality and accessibility of healthcare services.

DOI: https://doi.org/10.5281/zenodo.21357805

Spatio-Temporal Assessment of Rainfall Variability in Gadag District, Karnataka

Authors: Research Scholar Shri. Suresh Lamani, Assistant Professor Dr.M.B. Chalawadi

Abstract: Rainfall is the key hydro-meteorological determinant governing agricultural sustainability in the semi-arid regions of North Interior Karnataka. This paper presents a rigorous statistical appraisal of rainfall characteristics in Gadag district over a 35-year epoch (1991–2025) across seven taluks Gadag, Mundargi, Shirhatti, Ron, Nargund, Lakshmeshwar, and Gajendragad. The long-term mean annual rainfall of the district is established at 542.53 mm with a standard deviation of 103.77 mm and a Coefficient of Variation (CV) of 19.13%, representing moderate inter-annual variability. Seasonal partitioning demonstrates the dominance of the South-West (SW) monsoon, contributing 66.8% (362.48 mm) to the annual budget, followed by the North-East (NE) monsoon at 16.7% (90.40 mm). Spatial analysis indicates that Shirhatti receives the highest annual mean (586.74 mm), while Nargund is the most arid taluk (501.69 mm). According to the India Meteorological Department (IMD) drought indices, the 35-year period witnessed 6 excess years, 7 deficit years, and 22 normal years. These findings offer critical baselines for predictive water budgeting, drought preparedness, and regional cropping patterns.

DOI: https://doi.org/10.5281/zenodo.21358170

BioAttend: An Edge-AI Client-Side Biometric Attendance and Anti-Spoofing System

Authors: Abhishek Choudhary, Sharukh, Sazid Ansari, Rohit Kumar, Assistant Professor Gautam Tyagi

Abstract: Traditional automated attendance monitoring environments suffer heavily from manual entry friction, administrative vulnerabilities like buddy-punching, and catastrophic data breaches due to central cloud server biometric processing. This paper introduces BioAttend, a novel, open-source architectural solution designed to overcome these bottlenecks by moving the complete lifecycle of biometric capture, validation, mapping, and authentication directly onto the browser-side runtime environment. Operating completely within a zero-server framework, BioAttend utilizes client-side TensorFlow.js via the face-api.js abstraction to parse real-time video feeds locally. The system executes three specific neural pathways: an optimized SSD MobileNet V1 framework for facial region localization, a FaceLandmarks68Net model to continuously isolate 68 coordinate vectors, and a FaceRecognitionNet architecture optimized for extracting deep 128-dimensional floating-point identification vectors. To defeat active spoofing vectors, we present an execution engine parsing structural shift in the Eye Aspect Ratio (EAR). This engine tracks physical user blinking signatures down to a specific temporal sequence, ensuring authentication occurs only upon confirmation of actual human physiological activity. Multi-factor safety checks are established using localized cryptographic handshakes integrated directly with a serverless Google Apps Script infrastructure to process transactional one-time passwords (OTPs) and send instantaneous administrative compliance logs. Computational profiles collected across heterogeneous operating systems demonstrate validation loops finishing inside 45ms with recognition accuracies of 99.4%, showing that robust institutional security, low latency, and full regulatory privacy compliance can be maintained concurrently within sandboxed device parameters.

DOI: https://doi.org/10.5281/zenodo.21393448

AI-Powered Mental Health Risk Detection in Clinical Environments

Authors: Harsh Mishra, Shivam Kumar, Ayushman Singh, Mohd Faisal, Riyanshu Saini

Abstract: Mental disorders have emerged as one of the biggest challenges in the contemporary healthcare world, particularly amongst students and working professionals. Traditional mental health care programs usually suffer from several problems including unavailability of services, shortage of mental healthcare professionals, lack of timely assistance, and social stigma of visiting psychological clinics [1]. This study proposes an AI-based mental health risk detection and support system termed 'Doctorina' that has been designed and developed using the MERN stack framework and AI technologies in order to offer immediate mental health support and counseling assistance. The proposed system consists of a role-based system incorporating students, counselors, and administrators. The proposed platform provides AI-powered conversational support using Groq API and the Llama 3.3 70B models to offer real-time guidance for achieving mental well-being and emotional assistance. The platform also provides facilities for appointment booking, counselor management, authentication, and data management using JSON Web Tokens (JWT). The aim of this research is to design and develop a scalable and accessible mental health care platform which can help patients to deal with their emotional problems at early stages. In this paper, we mainly describe the system architecture, artificial intelligence integration, workflow, and implementation strategies for the development of the platform.

DOI: https://doi.org/10.5281/zenodo.21393647

Architecting Trust and Efficiency in Human Resources: A Comprehensive Integration of NLP Resume Parsing, Generative AI, and Retrieval-Augmented Generation (RAG)

Authors: Mohit Pundir

Abstract: The integration of Artificial Intelligence (AI) in Human Resources (HR) and talent ac-quisition has transitioned from an experimental phase to an operational necessity, driven primarily by the proliferation of Large Language Models (LLMs) and advanced machine learning algorithms. While AI introduces unprecedented efficiency in recruitment automa-tion, it simultaneously risks eroding organizational trust and the interpretability of HR systems. This paper presents a comprehensive, multi-layered approach to modernizing HR infrastructure without sacrificing human-centric decision-making. First, the paper exam-ines the macroeconomic impacts of AI-driven recruitment automation, highlighting shifts in hiring cycles, cost reductions, and global adoption trends. Second, an offline, Natural Language Processing (NLP) based resume parsing framework is proposed, utilizing tools such as spaCy and regex to structure applicant data efficiently, demonstrating an over-all accuracy of 96.0%. Finally, to mitigate the systemic risks of algorithmic black boxes and judgment inflation, the paper introduces Retrieval-Augmented Generation (RAG) as a foundational cybernetic design pattern for HR. By separating memory from generation, the RAG framework ensures that AI acts as an explainability interface and a sensemaking copilot rather than an autonomous decision-maker. Through this synthesis, the paper ar-gues that the ultimate goal of AI in HR is not merely the automation of transactions, but the architectural preservation of dignity, narrative consistency, and institutional memory.

DOI: https://doi.org/10.5281/zenodo.21394687

Autonomous AI-Based Triage System with Human Oversight” (MedTriage AI™)

Authors: Suryansh Anand, Mandeep, Dr. Vikalpa Tyagi

Abstract: High-volume clinical environments and remote patient monitoring pipelines are increasingly constrained by high cognitive loads, staffing shortages, and data saturation. While automated decision-making systems promise to optimize workflow efficiency, monolithic architectures introduce acute risks of automation bias, "explainability crises," and catastrophic undertriage of time-critical conditions. This paper introduces AegisTriage, a scalable, multi-agent autonomous artificial intelligence (AI) framework developed to dynamically assess patient acuity while integrating continuous, meaningful human oversight. Built using a distributed, tool-enabled Model Context Protocol (MCP), AegisTriage decomposes complex patient encounters into specialized sub-tasks managed by independent clinical agents: an Intake & History Agent, a Multimodal Context Synthesis Agent, and an Algorithmic Prioritization Agent. Rather than replacing the human clinician, the system operates as an intellectual filter that structures complex data, generates explicit clinical reasoning paths, and maintains human-in-the-loop (HITL) gates for high-acuity classifications. We evaluate the system using a retrospective cohort of simulated and real-world clinical vignettes against a human majority-vote gold standard. AegisTriage achieved a 95.8% sensitivity for emergency classifications and an 88.5% sensitivity for overall actionable alerts, operating at an average processing cost of

.34 per triage encounter. Crucially, the integration of explainable clinical reasoning and structured override protocols significantly lowered automation bias in human operators during simulated stress testing. These findings present a scalable, legally compliant pathway for augmenting high-stakes clinical and remote triage without sacrificing professional accountability or patient safety.

DOI: https://doi.org/10.5281/zenodo.21395214

Continual Learning Systems for Evolving Disease Patterns in Hospitals

Authors: Anshika Saini, Radha Kumari, Tanvi Verma, Vanshika Gupta, Assistant Professor Dr. Amit Kumar

Abstract: A silent but critical challenge confronts every hospital’s AI system after deployment: disease patterns shift, patient demographics evolve, and new outbreaks emerge, yet the underlying model remains frozen at its training snapshot. This temporal mismatch causes accuracy to erode steadily without any visible warning signals to clinical staff. The present work constructs and validates a Continual Learning (CL) pipeline specifically engineered to eliminate this vulnerability. Three coordinated mechanisms form the backbone of the approach. An Elastic Weight Consolidation (EWC) regularizer identifies which model parameters are most responsible for previously learned disease distinctions and applies a Fisher-Information-weighted penalty to prevent those parameters from drifting during future updates. A fixed-capacity Experience Replay buffer stores five hundred past patient records that are blended with each new data batch at training time, ensuring earlier disease representations are continuously refreshed. An ensemble of three independent drift sensors—ADWIN, DDM, and the Page-Hinkley test—runs concurrently on the incoming patient stream; the moment any sensor signals a statistically significant distributional change, an incremental retraining cycle is triggered automatically. Four clinical scenarios are simulated to exercise the system: a routine disease baseline, a sudden COVID-19 variant surge, a monsoon-driven dengue outbreak, and a post-outbreak mixed-prevalence recovery. Evaluated across these phases, the pipeline attains a mean classification accuracy of 88.1% over five diseases using ten routine clinical measurements. Most significantly, when Phase 1 data is re-evaluated after three complete learning cycles, accuracy falls by only 1.0%, confirming that knowledge acquired during initial training survives subsequent adaptation intact. Drift is detected at every phase boundary with no false alarms, and post-update accuracy recovers within a single retraining cycle. A Streamlit monitoring application delivers live accuracy trends, configurable alert thresholds, and drift event histories to clinical administrators without requiring any machine learning expertise.

DOI: https://doi.org/10.5281/zenodo.21405942

Deploykit-cli

Authors: Himanshu Haldar, Deepanshi Dhuria, Santosh Kumar, Adarsh Tripathi, Dr. Md Iqbal

Abstract: DeployKit is an open-source command-line tool that automates the full deployment of web applications on Ubuntu VPS servers. Unlike Docker-based PaaS solutions, it targets low-memory (512 MB+) environments, installing Node.js, Nginx, PM2, Let’s Encrypt SSL, and other components in one command. This paper analyzes DeployKit’s design, implementation, and performance compared to existing tools. We find that DeployKit significantly reduces setup time (~2 minutes) and resource requirements, making it well-suited for lightweight servers, at the cost of supporting fewer project types than broader PaaS systems. By operating directly on the native host operating system without container virtualization, DeployKit achieves high-efficiency deployments on servers with as little as 512MB of RAM. This report details the system architecture, smart project detection algorithms, automated memory-limiting strategies, and Nginx performance tuning of DeployKit.

DOI: https://doi.org/10.5281/zenodo.21406164

Design and Development of a Full-Stack Dashboard for Real-Time Bitcoin Portfolio Tracking and Candlestick Chart Analysis

Authors: Chetan, Adarsh Kushwaha, Nidhi Sinha, Sudhanshu Singh, Assistant Professor Dr.Amit Kumar

Abstract: Volatile price fluctuation and fragmented market intelligence present significant challenges for cryptocurrency investors attempting to maintain a balanced asset portfolio. To address these inefficiencies, this paper details the design and deployment of an integrated full-stack web ecosystem engineered for low latency portfolio orchestration, real-time bitcoin telemetry tracking and multi-interval candlestick visualization. Built as a decoupled architecture, the client interface leverages a dynamic react.js state engine to map live asset updates, while an asynchronous node.js backend manages data aggregation. The server infrastructure establishes stateful pipelines connecting to external markets and points – specifically the CoinGecko API spot-rate pricing and independent syndication networks for real time sector news delivery. User credential, transactional balances, and relational folding logs are permanently tracked using structural NOSQL schema within a MongoDB cluster. Rather than acting as a passive monitor, the platform functions as an interactive work space that aggregates historical asset variation, handles precision local currency mapping, and provides time-bucketed candlestick charts for financial pattern identification. Experimental deployment evolution indicates that the system successfully mitigates data latency constraints and eliminates user interface layout bottle neck the resulting platform delivers high throughput tracking and structured, single-pane data integrity capable of supporting rapid, evidence-based asset reallocation decisions.

DOI: https://doi.org/10.5281/zenodo.21406411

Emergency SOS & Live Location Tracking App

Authors: Anit, Divyanshu Nishad, Nikhil Kumar

Abstract: The rise in emergency situations — accidents, crimes, and medical crises — has made personal safety a critical concern. Existing communication methods often fail under high stress, resulting in delayed responses and increased risk. This paper presents Raksha360, a cross-platform mobile application that enables users to dispatch real-time SOS alerts with live GPS coordinates to trusted contacts via a single tap. Built with Flutter and Firebase Firestore, the system integrates the Geolocator package for high-accuracy GPS acquisition and WhatsApp/SMS for multi-channel alert delivery. Evaluation across multiple devices and network conditions achieves sub-5-second alert delivery, ≈5 m outdoor GPS accuracy, and 97% cloud write reliability, confirming the architecture's viability as a scalable, production-ready personal safety solution.

DOI: https://doi.org/10.5281/zenodo.21406801

Event-Driven Al Systems for Hospital Analytics

Authors: Ravi Ranjan, Sahil Kumar, Rohit Kumar, Aaryan Singh, Chirag Choudhary, Dr. Vinod Kumar

Abstract: The rapid digital transformation of healthcare organizations has significantly increased the adoption of cloud-based Enterprise Resource Planning (ERP) systems for managing financial, workforce, and supply-chain operations. Among these platforms, Workday has emerged as a leading solution due to its scalability, unified architecture, and real-time operational capabilities. However, traditional ERP systems are primarily rule-based and often fail to adapt dynamically to rapidly changing healthcare environments where operational events occur continuously and require immediate intelligent responses. Challenges such as inventory shortages, delayed payments, workforce imbalances, rising operational costs, and fluctuating patient demand highlight the limitations of conventional workflow automation methods. This research proposes an AI-enabled event-driven orchestration framework integrated within Workday ERP to improve operational efficiency and decision-making in healthcare organizations. The proposed framework combines machine learning techniques, predictive analytics, anomaly detection mechanisms, and process mining strategies to automate and optimize financial and supply-chain workflows across distributed healthcare systems. Unlike static ERP workflows, the proposed system continuously monitors operational events in real time and triggers intelligent automated actions based on contextual analysis and predictive insights. The framework utilizes event-driven architecture principles to synchronize multiple healthcare operations, including procurement management, inventory tracking, vendor coordination, payment processing, resource allocation, and workforce planning. AI-driven triggers are employed to identify abnormal operational conditions such as sudden inventory depletion, delayed invoice settlements, unusual purchasing patterns, and demand fluctuations. improved inventory accuracy, enhanced financial transparency, faster response times, and increased workflow synchronization. Furthermore, the integration of AI capabilities within Workday’s event-based ecosystem enhances scalability, operational resilience, and intelligent automation readiness for modern healthcare enterprises. The findings of this research contribute to the growing field of intelligent ERP systems by presenting a practical framework for AI-assisted workflow orchestration in healthcare environments. The study also establishes a foundation for future advancements in autonomous ERP ecosystems, predictive operational management, and next-generation healthcare automation strategies.

DOI: https://doi.org/10.5281/zenodo.21408778

Event-Driven Al Systems for Hospital Analytics

Authors: Ravi Ranjan, Sahil Kumar, Rohit Kumar, Aaryan Singh, Chirag Choudhary, Dr. Vinod Kumar

Abstract: The rapid digital transformation of healthcare organizations has significantly increased the adoption of cloud-based Enterprise Resource Planning (ERP) systems for managing financial, workforce, and supply-chain operations. Among these platforms, Workday has emerged as a leading solution due to its scalability, unified architecture, and real-time operational capabilities. However, traditional ERP systems are primarily rule-based and often fail to adapt dynamically to rapidly changing healthcare environments where operational events occur continuously and require immediate intelligent responses. Challenges such as inventory shortages, delayed payments, workforce imbalances, rising operational costs, and fluctuating patient demand highlight the limitations of conventional workflow automation methods. This research proposes an AI-enabled event-driven orchestration framework integrated within Workday ERP to improve operational efficiency and decision-making in healthcare organizations. The proposed framework combines machine learning techniques, predictive analytics, anomaly detection mechanisms, and process mining strategies to automate and optimize financial and supply-chain workflows across distributed healthcare systems. Unlike static ERP workflows, the proposed system continuously monitors operational events in real time and triggers intelligent automated actions based on contextual analysis and predictive insights. The framework utilizes event-driven architecture principles to synchronize multiple healthcare operations, including procurement management, inventory tracking, vendor coordination, payment processing, resource allocation, and workforce planning. AI-driven triggers are employed to identify abnormal operational conditions such as sudden inventory depletion, delayed invoice settlements, unusual purchasing patterns, and demand fluctuations. improved inventory accuracy, enhanced financial transparency, faster response times, and increased workflow synchronization. Furthermore, the integration of AI capabilities within Workday’s event-based ecosystem enhances scalability, operational resilience, and intelligent automation readiness for modern healthcare enterprises. The findings of this research contribute to the growing field of intelligent ERP systems by presenting a practical framework for AI-assisted workflow orchestration in healthcare environments. The study also establishes a foundation for future advancements in autonomous ERP ecosystems, predictive operational management, and next-generation healthcare automation strategies.

DOI: https://doi.org/10.5281/zenodo.21408778

Face Based Attendance System

Authors: Shrijana kumari Dixit, Ritik Sinha, Pratuish Anand, Saumya, Professor Dr. Md Iqbal

Abstract: Conventional attendance management systems rely heavily on manual recording, RFID-based identification, or contact-dependent biometric methods, which often suffer from issues including proxy attendance, operational delays, scalability limitations, and reduced analytical capabilities. To address these challenges, this research presents FaceAIDetect, an intelligent face recognition and emotion-aware attendance management framework that integrates Computer Vision, Artificial Intelligence, and Machine Learning for automated identity verification and attendance generation. The proposed system performs real-time facial acquisition, preprocessing, face detection, feature extraction, embedding generation, similarity-based matching, emotion recognition, and automated attendance recording within a unified architecture. Facial representations are generated through encoding mechanisms and indexed using similarity search techniques to improve recognition speed and scalability. The framework extends traditional attendance systems by incorporating emotional state analysis, enabling contextual understanding of user behavior during attendance operations. The implementation utilizes Python-based technologies including OpenCV, FAISS, MongoDB, and modern machine learning pipelines for recognition and storage management. Experimental evaluation demonstrates improvements in operational efficiency, reduced manual intervention, enhanced attendance authenticity, and effective real-time performance under controlled deployment conditions. The proposed system contributes toward the development of intelligent and contactless attendance infrastructures suitable for educational institutions, workplaces, and secure monitoring environments. Future enhancements may include cloud deployment, federated learning integration, mobile accessibility, and advanced behavioral analytics.

DOI: https://doi.org/10.5281/zenodo.21410481

Full Stack Dashboard for Real-Time Data Visualization

Authors: Gaurav Kumar, Khushboo Yadav, Bansi Saini, Assistant Professor Miss Rimmy

Abstract: A real-time data visualization dashboard is a system that enables users to monitor and analyze data instantly through graphical representations such as charts and graphs. It helps organizations and individuals make faster and more accurate decisions by presenting live data in an easy-to-understand format. In this project, a secure and scalable real-time dashboard has been developed using the MERN Stack (MongoDB, Express.js, React.js, and Node.js). The system includes features such as dynamic chart generation, dataset upload, real-time updates, and user authentication. Socket.IO is used to enable low-latency communication between the server and client, allowing instant updates without page reload. MongoDB ensures efficient storage and retrieval of large datasets, while JWT-based authentication secures user access. This research presents an effective approach for building modern, scalable, and real-time data visualization systems. Future enhancements may include AI-based analytics, predictive insights, and advanced reporting features.

DOI: https://doi.org/10.5281/zenodo.21410894

Full Stack Project Management Tool with AI Integration

Authors: Shashank Awasthi, Shivang Kumar, Anshika Negi, Ms. Nisha Sharma

Abstract: Modern projects are really complicated which makes it very hard for teams to plan, schedule and keep track of things. Traditional tools for managing projects are good for assigning tasks and tracking progress. They do not do a good job of helping with important things like figuring out which tasks are most important making sure everyone has a fair amount of work predicting when things will be done, identifying potential problems and making reports automatically. This paper is about a project management tool that uses Artificial Intelligence and a set of technologies including MongoDB, Express.js, React.js and Node.js to create a smart and interactive platform. The system takes in information about a project, such as what the goals are, who is on the team when things are due and what tasks need to be done and then it automatically creates a detailed plan, a realistic schedule, a map of how tasks are related a fair distribution of work warnings about potential problems and a summary of progress. By combining a foundation with a smart Artificial Intelligence engine the tool reduces the amount of manual work that needs to be done and it helps teams make better decisions and be more productive. Testing the system and getting feedback from users shows that it makes a difference, in how fast plans can be made how well tasks are organized and how happy users are. The new system shows how Artificial Intelligence can be used to make project management better. It takes ideas that're not well organized and turns them into clear plans that can be put into action.

DOI: https://doi.org/10.5281/zenodo.21412003

Assessing Foundational Literacy and Numeracy Outcomes in Haryana: A Study of Class IV Hindi and Mathematics Achievement under the NIPUN Bharat Mission

Authors: Research Scholar Ms. Nidhi Sharma, Professor Dr. Raj Kumar

Abstract: Foundational Literacy and Numeracy (FLN) is a flagship national mission launched under India’s National Education Policy (NEP) 2020 and implemented through the NIPUN Bharat programme. Its core objective is to ensure that every child attains grade-level proficiency in basic reading, writing, and arithmetic—including letter and number recognition, reading with comprehension, sentence construction, and foundational operations—by the end of Grade III, recognizing these skills as the gateway to all future learning. This study assessed FLN achievement among 800 Class IV students across 40 government schools in four districts of Haryana—Sirsa, Panchkula, Faridabad, and Jhajjar—using standardized FLN-aligned tests in Hindi and Mathematics. Quantitative analysis through means, SDs, t-tests, and ANOVA revealed a clear pattern: foundational skills are well established, with 90–100% achievement in Varna and Number Recognition. However, higher-order competencies remain moderate, with comprehension, sentence writing, subtraction, and multiplication ranging from 35% to 48.8%. Overall, the findings indicate that FLN interventions have successfully strengthened decoding and number sense, yet universal proficiency is still incomplete. Achieving NEP 2020 goals will require focused remediation in comprehension, writing, and advanced arithmetic, along with strategies to address existing gaps.

DOI: https://doi.org/10.5281/zenodo.21425162

Gen AI Driven Hospital Automation Framework

Authors: Wasiur Rahman, Mohd Rahman

Abstract: Healthcare systems are rapidly evolving with the integration of Artificial Intelligence and intelligent automation technologies. Traditional hospital management systems often face challenges such as excessive paperwork, delayed patient services, inefficient resource utilization, data management issues, and lack of real-time decision-making capabilities. To overcome these limitations, this research proposes a Gen AI Driven Hospital Automation Framework that utilizes Generative Artificial Intelligence to automate and optimize hospital operations efficiently and securely. The proposed framework integrates advanced technologies such as Generative AI, Machine Learning (ML), Natural Language Processing (NLP), Cloud Computing, Internet of Things (IoT), predictive analytics, and database management systems to create an intelligent healthcare ecosystem. The system automates key hospital functions including patient registration, appointment scheduling, electronic health record management, AI-assisted diagnosis, pharmacy and inventory management, billing systems, emergency response services, and healthcare monitoring. The framework employs AI-powered virtual assistants and chatbots to provide real-time patient communication, symptom analysis, appointment support, and automated healthcare guidance. Machine Learning algorithms analyze patient data, medical history, and clinical reports to assist doctors in accurate diagnosis and treatment recommendations. Additionally, cloud-based healthcare infrastructure enables secure data storage, real-time accessibility, remote healthcare services, and efficient coordination between hospital departments. The proposed system also focuses on healthcare security and privacy by implementing encrypted cloud storage, role-based access control, authentication mechanisms, and secure patient record management. Automation of repetitive administrative tasks reduces human errors, minimizes operational costs, improves healthcare efficiency, and allows medical staff to focus more on patient care. Experimental analysis and system evaluation demonstrate that the proposed framework significantly improves hospital performance by reducing patient waiting time, increasing diagnosis accuracy, enhancing communication efficiency, optimizing resource management, and improving patient satisfaction. The intelligent automation capabilities of the system contribute toward building smart hospitals capable of delivering faster, safer, scalable, and personalized healthcare services. This research highlights the transformative potential of Generative AI in modern healthcare and contributes toward the development of intelligent next-generation hospital automation systems.

DOI: https://doi.org/10.5281/zenodo.21426457

Full-Stack AI-Powered E-Commerce Platform

Authors: Abhishek Kumar, Astha, Kriti Kumari, Shagun Pundir, Assistant Professor Miss Jaishree Goyal

Abstract: The rapid growth of online retail has pushed the demand for smarter, more adaptive, and user-focused shopping systems to new heights. Most platforms available today still depend on outdated keyword matching and rigid, rule-based filtering, which often fall short when a shopper phrases their intent in natural language or when their needs evolve mid- session. This paper describes the design and practical implementation of a full-stack e-commerce platform that weaves semantic search, secure payment handling, and intelligent order management into one cohesive system. At its heart, the platform uses the Gemini API for transformer-based language understanding, allowing it to interpret what a user actually means—not just what they typed. Stop-word removal and dynamic query construction further sharpen retrieval accuracy. The backend runs on Node.js and Express, with PostgreSQL providing relational data integrity and transactional reliability. User authentication is handled through JWT-based tokens, payments flow through Stripe, and product images are stored via Cloudinary. Evaluation results show measurable gains in search precision and user engagement over traditional approaches, with computational overhead remaining well within acceptable bounds. Taken together, the findings confirm that embedding semantic intelligence into an e-commerce stack meaningfully improves product discoverability, operational efficiency, and the overall shopping experience.

DOI: https://doi.org/10.5281/zenodo.21426575

GEN-AI Driven Automated Medical Report Generation and Discharge Summary

Authors: Devansh, Darshit Chauhan, Aditya Singh, Assistant Professor Mr. Anurag Chandana

Abstract: The healthcare industry generates enormous volumes of clinical data daily, including patient histories, diagnostic results, lab reports, and physician notes. Converting this unstructured data into structured, readable medical reports and discharge summaries is a time-consuming and error-prone process when done manually. This research presents a GEN-AI driven system for automated medical report generation and discharge summary creation using large language models (LLMs), machine learning (ML) algorithms, and natural language processing (NLP) techniques. The system accepts raw clinical inputs including patient demographics, symptoms, diagnosis codes, prescribed medications, and lab results and automatically generates structured medical reports and professional discharge summaries. The project utilizes a web-based interface developed with modern frontend technologies and integrates backend ML pipelines for intelligent content generation. The results demonstrate significant improvements in documentation speed, accuracy, and consistency compared to manual approaches.

DOI: https://doi.org/10.5281/zenodo.21427004