Volume 14 Issue2

Alast Circle: A Centralized Student And Alumni Interaction System

Authors: Harsh B Sheth

Abstract: Alast circle is a centralized school and university community platform designed to maintain continuous interaction between administrators, staff members, students, and alumni. In many institutions, communication with students becomes limited after graduation, and important announcements or events are often shared through scattered channels. The proposed system addresses this issue by providing a single web-based platform where current students and former students can remain connected with their institution. Alast circle enables users to receive announcements, participate in events, and communicate with one another within an organized environment. The platform also allows administrators and staff to manage updates, publish event details, and maintain institutional engagement. By integrating communication, event participation, and user interaction into one system, Alast circle strengthens long-term relationships within the academic community and improves information accessibility for all users.

DOI: https://doi.org/10.5281/zenodo.19664046

Iot-Enabled Advanced Water Quality Monitoring System and Analysis of Water Consumption Using Flow Sensor

Authors: Mr. Satheesh Kumar D, Mr. Kishore Kumar S, Mr. Mohammed Abdul Jaleel J, Mr. Praveen Kumar S, Mr. Perumal S

Abstract: Water is one of the most vital resources for human survival, and maintaining quality and proper usage is a growing concern in modern urban environments. Corporation water supplied to residential and commercial buildings often experiences contamination due to pipeline corrosion or improper treatment. To ensure safe and efficient water usage, this project proposes an IoT-based water quality and usage monitoring system. The system integrates multiple sensors pH sensor, a TDS sensor, a temperature sensor, and a retrofit meter to continuously monitor the chemical and physical characteristics of water. The data collected from these sensors is processed by an Arduino microcontroller and transmitted over Wi-Fi to a cloud server or dashboard for real-time visualization and analysis. This helps detect contamination, water wastage, or abnormal consumption patterns early. The proposed system provides a low-cost, scalable, and automated solution suitable for smart city infrastructure, corporate facilities, and residential water distribution networks. By enabling continuous monitoring and data-driven decision-making, the system ensures water safety, supports conservation efforts, and promotes sustainable water management.

DOI: https://doi.org/10.5281/zenodo.19669642

Design And Implementation Of Secure Wireless Network Deployment With SRX300

Authors: Mr. Satheeshkumar S, Mrs.Thendral K, Madhan Kumar S, Madhan R, Mukesh Pandian A, Ravivarma R

Abstract: Wireless networking has become a fundamental component of modern organizations and educational institutions, enabling users to access the internet and communicate efficiently. However, these networks are often vulnerable to several security challenges, including unauthorized device access and the presence of rogue access points. This work focuses on the design and deployment of a secure wireless network using Juniper networking technologies such as the SRX300 firewall, EX series switches, and Mist AP32/AP63 wireless access points. Network segmentation is implemented using Virtual Local Area Networks (VLANs) to separate traffic and enhance overall network security and management. A dedicated guest wireless network is also configured through the Juniper Mist Cloud platform using a captive portal authentication mechanism, which provides controlled access for external users. The SRX300 firewall is used to enforce security policies that regulate network traffic and restrict unauthorized activities. Furthermore, advanced security features including rogue access point detection is integrated to identify and analyze potential wireless threats. The developed system demonstrates an effective solution for establishing a secure, scalable, and well-managed enterprise wireless network environment.

DOI: https://doi.org/10.5281/zenodo.19670690

Cloud Computing: Architectural Frameworks, Security Considerations, And Advanced Technologies

Authors: Mrs Shwetambari Sukhadeo Dhupe

Abstract: Cloud computing has emerged as a dominant computing paradigm by enabling flexible, scalable, and cost-effective access to shared computing resources over the internet. This research paper presents a comprehensive examination of cloud computing architectures, service and deployment models, security and privacy mechanisms, performance enhancement strategies, and emerging technological trends. The study highlights key challenges such as data protection, interoperability, and regulatory compliance, while also exploring future research directions including edge computing, serverless models, and sustainable cloud infrastructures.

Outdoor Wifi Coverage & Environmental Optimization With Juniper AP63

Authors: Dr.Thirumala Lakshmi K, Mathan S, Kavin S, Praveen Kumar P, Rishi S, Mrs. Thendral K

Abstract: In the era of smart campuses, seamless outdoor wireless connectivity is essential for enabling real-time applications, IoT integration, and uninterrupted user experiences. This project focuses on extending and optimizing outdoor Wi Fi coverage using Juniper AP63 access points, supported by Mist AI for intelligent performance monitoring and environmental adaptation. The deployment aims to eliminate dead zones and ensure robust signal transitions between outdoor AP63 and indoor AP32 zones through seamless device handoff. To address the unique challenges of outdoor environments such as signal degradation from terrain, foliage, and weather this study integrates insights from advanced literature, including clustering-based AP placement, energy-efficient AP activation, and RF propagation modeling. The project also explores environmental optimization strategies to enhance coverage reliability and reduce energy consumption. Performance validation is conducted through field testing and simulation, ensuring that the proposed solution meets the demands of a dynamic, high-density campus environment. This work contributes to the development of scalable, AI-driven outdoor wireless networks that are resilient, energy-aware, and capable of supporting future smart campus innovations.

DOI: https://doi.org/10.5281/zenodo.19671607

Implementation of Vlan, Ospf and Inter-Vlan Routing for Seamless Connectivity

Authors: Ms.Thendral K, Kaviya Sri K, Padmapriya B, Parameshwari P, Rubashri M

Abstract: In modern enterprise and campus networks, maintaining seamless connectivity, scalability, and performance is a major challenge when multiple departments share the same network segment. Such configurations often lead to excessive broadcast traffic, reduced efficiency, and security vulnerabilities. Additionally, static routing becomes inefficient and prone to configuration errors during network changes. This project focuses on implementing Virtual Local Area Networks (VLANs) and OSPF-based Inter-VLAN Routing on Juniper SRX300 devices to achieve optimized network segmentation and dynamic routing. VLANs are configured to logically separate departmental traffic, thereby reducing broadcast domains and enhancing network security. Inter-VLAN routing enables communication between different VLANs, ensuring seamless connectivity across departments. The Open Shortest Path First (OSPF) protocol is used to provide scalable and adaptive routing, allowing the network to automatically adjust to topology changes. The iPerf tool is utilized to analyze network performance parameters such as throughput, latency, and bandwidth utilization, validating the effectiveness of the proposed setup. Experimental results demonstrate improved network performance, reduced broadcast traffic, and enhanced routing efficiency, ensuring a secure, reliable, and high-performance network infrastructure.

DOI: https://doi.org/10.5281/zenodo.19675180

AI Based Travel Itinerary System

Authors: Amol Pramod Parab, Yash Kishor Nadkar, Vighnesh Sanjay Sathe, Prof. Sanjay N. Jadhav

Abstract: The AI-Based Travel Itinerary System is an intel- ligent web application designed to simplify trip planning by automatically generating personalized travel itineraries based on user preferences. The system uses Artificial Intelligence to analyze inputs such as destination, travel duration, budget, and interests. It integrates APIs like Google Maps and AI models to provide real-time recommendations including routes, attractions, and accommodations. The system reduces manual effort and enhances travel planning efficiency by generating optimized schedules and cost estimations.

DOI: https://doi.org/10.5281/zenodo.19675227

Smart Vision Care System Using Ai Based Eye Activity Recognition Air Blower Control

Authors: Mrs. A.K. Nivedha M.E., S.Pappathi, J. Priyadharshini, K.Sivapriya, B. Subasini

Abstract: The rapid increase in digital device usage has led to significant eye-related problems such as eye strain, dryness, and reduced blinking rate. Continuous exposure to screens without proper eye care can negatively impact human vision and overall health. To address this issue, this project proposes an AI-based Smart Vision Care System that monitors eye activity in real time and provides automatic relief mechanisms. The proposed system utilizes a camera to continuously capture the user’s eye movements and applies image processing techniques using artificial intelligence to detect eye states such as blinking frequency and eye closure duration. By analyzing these parameters, the system can identify signs of eye fatigue or strain. When abnormal eye activity is detected, the system automatically activates an air blower to provide cooling and relaxation to the eyes without requiring any manual intervention. The system is developed using software tools such as Python and OpenCV for eye detection and recognition, along with hardware components including a microcontroller, camera module, and air blower. This integration of AI and embedded systems ensures accurate monitoring and timely response. The main objective of this project is to reduce eye strain, improve user comfort, and promote healthy screen usage habits. The proposed system is cost-effective, efficient, and suitable for students and professionals who spend long hours in front of digital devices. In the future, the system can be enhanced by integrating alert notifications and mobile applications for better user interaction and monitoring.

Loan Approval Prediction System

Authors: S.Rohith Reddy, P.Sri Nanda Kishore, Dr.R.sivaramakrishnan

Abstract: The Loan Approval Prediction System is a machine learning-based application developed to automate the process of evaluating loan applications. It analyzes applicant data such as income, credit history, employment status, and loan amount to predict whether a loan should be approved or not. By using historical data and applying data preprocessing techniques, the system improves the accuracy and reliability of predictions. The model is built using classification algorithms such as Logistic Regression, Decision Tree, or Random Forest, and the best-performing model is selected for deployment. This system helps financial institutions reduce manual effort, minimize risk, and make faster, data-driven decisions. It also enhances efficiency and ensures a more consistent and unbiased loan approval process. Overall, The Loan Approval Prediction System is a machine learning-based application that predicts whether a loan will be approved based on user details like income and credit history. It uses models such as Logistic Regression, Random Forest, and XGBoost for accurate decision-making. Built with Python and Flask, it provides quick, data-driven loan predictions through a simple web interface.

Smart Flood Water Evacuation System In Urban Areas

Authors: Abinaya S, Jayasoorya S, Lathika K, Dr. B. Veerasamy, Dr. G. Gandhimathi

Abstract: Urban flooding has become a recurring and critical challenge due to rapid urbanization, inadequate drainage infrastructure, and unpredictable climate conditions. This paper presents an enhanced Smart Flood Water Evacuation System based on an ESP32 (NodeMCU-32) IoT platform integrated with analog float sensors and a three-tank sequential water routing mechanism. The proposed system continuously monitors water levels in three independent storage tanks using calibrated analog sensors with a 15-sample averaging filter for noise-free readings. An automatic relay-based pump control activates or deactivates each motor based on predefined threshold levels (safe: 30%, warning: 60%, critical: forced ON). Floodwater is intelligently routed through Tank 1, Tank 2, and Tank 3 sequentially before being diverted to the outlet, enabling water conservation and reuse. Real-time sensor data and system status are published to a HiveMQ cloud-based MQTT broker over a secure TLS connection, enabling remote monitoring and manual override via any MQTT-compatible dashboard. An integrated battery management module automatically controls the charging relay based on voltage thresholds, ensuring sustainable solar-powered operation. Experimental results demonstrate high detection accuracy, sub-second MQTT latency, and reliable autonomous operation. The proposed system is low-cost, scalable, and practically deployable in urban flood-prone areas.

DOI: https://doi.org/10.5281/zenodo.19675764

Neural Net And Deep Learning Based Brain Lump Categorization & Intensity Gradient Platform

Authors: Prashant Yadav, Mohd Danish, Md Zishan Ansari

Abstract: Early and reliable detection of Brain lumps including mening and gliome, remains a significant concern in clinical practice. Conventional Magnetic Resonance Imaging (MRI) analysis relies heavily on expert interpretation, which may introduce variability and delay diagnosis. This study describes an integrated deep learning-based system for automated brain tumor detection. The developed system, termed the Automated Neuro-Diagnostic Assistant (ANDA), is designed using a customized Convolutional Neural Network (CNN) trained on preprocessed MRI datasets. A key feature of the system is its deployment as an interactive real-time web application using Flask, incorporating modules such as a confidence score visualizer and an automated report generator. Experimental results indicate a classification accuracy of 96% along with an F1-score of 94%, demonstrating reliable prediction capability. The proposed evaluation metric (Fscore) provides a unified assessment of system performance by combining accuracy, interpretability, and usability factors. Fscore = ⌊0.35A + 0.25C + 0.15V + 0.10(HI + SI + UI)⌉ Where A represents classification accuracy, C denotes confidence reliability, V indicates the effectiveness of visual interpretation, HI refers to healthcare insight generation, SI represents stroke identification capability, and UI denotes overall system usability. Further evaluation on diverse MRI samples shows strong agreement (r = 0.89) with expected diagnostic patterns and consistent performance across tumor categories. The system improves interpretability and user interaction efficiency, enabling faster and more structured medical image analysis. Overall, the system provides stable processing, real-time prediction capability, and efficient handling of medical data. By integrating automated detection with supportive visual outputs, it enhances accessibility to AI-assisted diagnostic tools and supports preliminary medical assessment.

Automatic Multi-Functional Agricultural Robot For Smart Crop

Authors: Aiswarya Sreejith, M.Radha, G.Subaparameshwari, Mrs.S.Revathi

Abstract: Agriculture is rapidly transforming with the integration of automation and intelligent systems. However, small and marginal farmers still rely heavily on manual labour for irrigation, pesticide spraying and crop monitoring. This work presents the design and development of an Automatic Multi-Functional Agricultural Robot intended for smart crop. The robot is capable of monitoring soil moisture, controlling irrigation automatically, spraying pesticides and navigating fields autonomously. By integrating sensors, a microcontroller and IoT-based monitoring, the system reduces human effort, optimizes water usage and enhances crop productivity. The system is affordable, energy-efficient and suitable for real-time agricultural applications.

DOI: https://doi.org/10.5281/zenodo.19675985

Smart Multimodal Neonatal Screening System Using Salivary Biomarkers

Authors: A. Naveen, N. Shahul Hameed, R. Thilagar, S. Vishwabharathi, Dr. S. Sutha

Abstract: Neonatal healthcare is a critical domain in modern medicine, focusing on the survival, growth, and development of newborn infants during the first few weeks of life. Early detection of neonatal disorders such as jaundice, dehydration, and neurological abnormalities plays a vital role in preventing long-term complications and reducing infant mortality rates. However, conventional diagnostic techniques primarily rely on blood-based tests, which are invasive, time-consuming, and often require well-equipped laboratory facilities and trained personnel. These limitations make such methods less suitable for continuous monitoring and challenging to implement in rural or low-resource settings. In recent years, there has been growing interest in the development of non-invasive diagnostic approaches that can provide accurate and real-time health assessment without causing discomfort to the patient. Among various biological fluids, saliva has emerged as a promising diagnostic medium due to its ease of collection, safety, and ability to reflect physiological and biochemical conditions of the body. Saliva contains a wide range of biomarkers, including enzymes, proteins, hormones, and metabolites, which can be used to detect various health conditions. The non-invasive nature of saliva collection is particularly beneficial for neonatal applications, where minimizing pain and stress is of utmost importance. The proposed system introduces a smart multimodal neonatal screening approach that utilizes salivary biomarkers for disease detection. By integrating microfluidic technology, optical sensing, and artificial intelligence, the system aims to provide a comprehensive and efficient diagnostic solution. Microfluidic chips enable precise handling of small sample volumes and facilitate controlled biochemical reactions, while optical sensors detect colour changes corresponding to biomarker concentrations. These signals are processed using embedded systems and analysed using machine learning algorithms to classify neonatal conditions accurately. Furthermore, the incorporation of smartphone-based imaging and mobile applications enhances the accessibility and usability of the system. The ability to store and transmit data through cloud platforms enables remote monitoring and telemedicine applications, making the system highly suitable for deployment in rural and underserved areas. Overall, the proposed system represents a significant advancement in neonatal healthcare by combining non-invasive diagnostics with modern technological innovations.

Design And Development Of A Low-Power Embedded Edge-Computing Framework For Real-Time Wildlife Monitoring and Deterrence

Authors: Mr.Loganathan S, Mr.Mohan R, Mr.Nitheesh K M, , Mr.Ramesh C, Mr.Saleem Ulla Khan S

Abstract: In modern agriculture, protecting crops from animal intrusions is a major challenge. This project presents a real-time wildlife monitoring and deterrence system using the YOLO V8 object detection algorithm. The system employs AI-based image processing with OpenCV for pre processing and integrates automatic notification and control mechanisms for enhanced farm security. A camera continuously captures images, and YOLO V8 detects and classifies animals in real time. Detected images are uploaded to a remote server for analysis and then deleted to save storage. Pre-processing steps like noise reduction, resizing, and normalization improve detection accuracy, while compression and feature extraction ensure real-time performance. When an animal is detected, the system sends an email alert with the timestamp and type of animal, activates a buzzer, and displays details on an LCD screen. LED floodlights turn on in low light to increase visibility and deter nocturnal animals. The YOLO V8 model is continuously refined for accuracy and adaptability, offering a practical, efficient solution for smart farm wildlife monitoring and deterrence.

DOI: https://doi.org/10.5281/zenodo.19676429

VidyaRaksha AI: A Quality Assurance Approach For Intelligent Online Assessment Systems

Authors: Diya Parmar, Dr. Chetan Soni

Abstract: With the growing shift toward digital education, the need for efficient and reliable online assessment systems has become increasingly important. Many existing platforms face challenges such as inaccurate evaluation, delayed results, and system reliability issues. This paper presents VidyaRaksha AI, an intelligent web-based assessment system designed to automate the examination process using Artificial Intelligence. The system supports key functionalities such as test creation, user participation, auto-submission, and instant result generation, which help reduce manual effort and improve overall efficiency. From a Quality Assurance (QA) perspective, the system was tested using techniques such as functional testing and regression testing to ensure correct system behavior and stability. Various test cases, including both positive and negative scenarios, were executed to identify issues like auto- submission errors, missing UI elements, and functional inconsistencies. These defects were reported and resolved to enhance system performance. The results indicate that the system provides accurate evaluation, faster result processing, and improved user experience. Overall, VidyaRaksha AI demonstrates a reliable and efficient approach to modernizing online assessment systems through the integration of AI and effective QA practices.

DOI: https://doi.org/10.5281/zenodo.19676606

Intelliquiz — A Gamified Adaptive Quiz Generator Using Llms with Real-Time Analytics

Authors: Avvari Pavithra, Y Jeevan Nagendra Kumar, Yenna Gnaneshwar Reddy, Ponnamanda Eswar, Madupathi Nithin Kumar

Abstract: Honestly, I’ve seen that most schools are still based on blunt tests and even though they make the exams are the same whether it is a person who is answering it but at his own pace, it’s a bad thing when people learn at different pace. One tool taking the game far further is IntelliQuiz which automatically generates entirely new quizzes using AI and sprinkles in some gamifying fun as well to keep things as interactive as possible. Instead of offering one format, it will take material from PDFs, videos and even just plain text, to construct its questions that shift in real time during class, based on how each student is doing. Harder questions pop up after a correct answer the easier ones pop up if someone’s struggling. Behind the curtain, the system has a combination of Spring Boot to manage server-related operations, React with Tailwind for frontend styling and storing everything in a PostgreSQL database. Access control divides the users into three groups namely Admin, Teacher and Student as per which they are provided only the permission as secured tokens. The engine really stands out because it fits large language models with retrieval methods to bring facts in into the room before generating the questions so that any prompt will be linked back to specific learning objectives and kept on point. The challenge scales over the levels of beginners, intermediaries and experts automatically without additional effort from teachers. It even tweaks questions on the fly based on how fast a learner is answering, the success rate and how his summers are without having to work on the testing questions instead of fixed testing questions, round by round helps to form itself around what a user actually knows. Points, badges and leaderboards pop-up as the progress is made encouraging that little quiet competition that keeps the attention wired. Live charts are updated as the activity comes through illustrating the gaps, the strengths and patterns. One benefit of this is that teachers may recognize trends early and students may realize where they want to expend their energy. The modular aspect to this means it can scale up very easily and shift very easily from one device to another including network condition. Every answer generated by AI gets screened first: it gets double checked before making its way to a student meanwhile the logic behind each score is traceable with difficulty weighed and usefulness measured. Regardless of whether the solution is online or offline, on the school servers or in a University data center, the solution remains lightweight. The time which until used to manually build quizzes now gets freed up via smart automation. Grading occurs plainly and obviously, comments are specific, and monitoring is close always using gripping levels of tracking. As validation is built in along the way, the trust is intact as the system becomes faster. Learning isn’t simply tracked it changes itself in real-time according to personal rhythm, pace, depth and direction. It’s not some type of magic or noise, it’s a constant consistency between effort and growth.

AI Based Interview Preparation Platform

Authors: K.Balaji, T.Toneswarreddy, S.Ramesh, P.Nihkil, N.Sannajaji

Abstract: The AI-Based Interview Preparation Platform is also created to facilitate a holistic, interactive and safe medium to candidates undergoing a job interview preparation. The site uses AI-powered simulation to create realistic interview experiences, providing real-time feedback and custom suggestions to users. The chatbot feature will allow users to train on a large set of interview questions, as well as to receive live and artificial intelligence-based answers, improving their training process. Also, the site uses a resume analysis which ultimately examines resumes according to Applicant Tracking System (ATS) standards by applying Natural Language Processing (NLP). This feature creates an ATS score and offers a lot of feedback so users can maximize their resumes and enhance their probability of attention by employers. The users are able to post their resumes, provide an understanding of the points that they need to improve, and adjust their content to the needs of a particular industry. The system also monitors user progress and suggests personalized interview questions according to their profile and the job specifications. It allows candidates to perfect their speaking and improve the answers and overall performance during simulated interviews as AI-derived feedback on the response assists them in honing their skills and improving their response. Moreover, the platform provides secure and confidential data to users, as well as uncomplicated navigation and scheduling of mock interviews. Combining NLP-driven resume analysis and AI-driven chatbot functionality, the platform provides an interactive and data-driven experience, helping job seekers to prepare for interviews better and shine in a competitive job market.

A Comparative Evaluation Of Machine Learning Algorithms For Early Detection Of Type 2 Diabetes

Authors: Seema Ahirwar, Rupali Chaure

Abstract: This paper presents a comparative evaluation of seven supervised ML classifiers Logistic Regression, Naive Bayes, k-Nearest Neighbors, Decision Tree, Random Forest, XGBoost, and SVM for early Type 2 diabetes (T2DM) prediction. Using the Pima Indians Diabetes Database (PIDD, n=768) and Frankfurt Hospital Diabetes Dataset (FHDD, n=2000), we apply standardized preprocessing with SMOTE-based class imbalance correction and stratified 10-fold cross-validation[1]. Models are evaluated on Accuracy, Precision, Recall, F1-Score, ROC-AUC, and MCC. Results show XGBoost consistently outperforms all classifiers (AUC: 0.901 PIDD, 0.951 FHDD), while Logistic Regression retains interpretability advantages for clinical deployment. Feature importance analysis identifies fasting plasma glucose, BMI, and HbA1c as top predictors, aligning with clinical guidelines.

DOI: https://doi.org/10.5281/zenodo.19678064

A Comprehensive Review Of Artificial Intelligence Techniques In Computer Science

Authors: Vanita Ganesh Bagal

Abstract: Artificial Intelligence (AI) has emerged as a foundational pillar of modern computer science, profoundly influencing a wide range of domains including automation, data analytics, natural language understanding, intelligent systems, and human–computer interaction. Rapid advancements in computational power, availability of large-scale datasets, and algorithmic innovations have accelerated the development and adoption of AI-based techniques across academic research and real-world applications. This review paper presents a comprehensive analysis of contemporary Artificial Intelligence techniques used in computer science, with particular emphasis on machine learning, deep learning, natural language processing, computer vision, reinforcement learning, generative models, and explainable artificial intelligence. The study systematically examines the computational foundations of these techniques, their underlying algorithms, and their practical applications in areas such as intelligent decision-making, pattern recognition, autonomous systems, and data-driven modeling. In addition, the paper highlights the advantages of AI techniques in improving efficiency, accuracy, and scalability of computing systems, while also discussing key limitations including model interpretability, data dependency, computational complexity, ethical concerns, and security challenges. Recent developments such as transformer-based architectures, large language models, and explainable AI frameworks are reviewed to illustrate current research trends and emerging directions. By synthesizing findings from recent surveys, scholarly articles, and systematic reviews, this paper provides an up-to-date overview of the state-of-the-art in Artificial Intelligence research within computer science. The review aims to serve as a valuable reference for researchers, academicians, and practitioners by identifying research gaps, outlining future opportunities, and emphasizing the need for responsible, transparent, and efficient AI systems in next-generation computing environments.

AI Interview Preparation System

Authors: Guntipalli Dhana Lakshmi, Kotha Harika, Kankipati John, Padala Srinivas

Abstract: The AI interview preparation system is a tool that uses automated evaluation and customized feedback to assist users in preparing for interviews. The system uses speech processing, keyword- based analysis, and LLM Modules to mimic interviewers. It creates questionnaires according to the domain and degree of difficulty chosen by the user. Similarity matching and scoring methods were used to assess the responses. Comments on accuracy, assurance, and clarity to assess communication effectiveness and speech input were also examined. It reduces the need for conventional interviewing techniques and gives users the opportunity to improve their performance through tracking and feedback mechanisms.

Enhancing Personal Finance Management Using AIBased Predictive Analytics And Intelligent Chatbot System

Authors: Anurag Mall, Nitin, Kirti Rastogi, Arvind Kumar

Abstract: The rapid growth of digital financial transactions, including UPI, online payments, and subscription-based services, has significantly increased the complexity of personal finance management. Despite the availability of various financial tracking applications, most systems lack intelligent analytics, predictive capabilities, and personalized guidance. This paper presents an AI-driven platform for enhancing Optimizing personal financial management using predictive analytics and an intelligent chatbot system. The proposed system integrates deterministic data handled using evaluation methods based on large language models to produce a multi-dimensional financial insight score. User financial data is processed through a structured pipeline, where expenses are categorized using machine learning algorithms, and spending behaviour is analysed using time-series predictive models. The system leverages a Natural Language Processing (NLP)-based chatbot that Allows users to query financial insights in instantly. A composite financial health score is computed using a weighted model: 𝑭𝒔𝒄𝒐𝒓𝒆 = ⌊𝟎. 𝟑𝟎𝑬 + 𝟎. 𝟐𝟓𝑷 + 𝟎. 𝟏𝟓𝑩 + 𝟎. 𝟏𝟎(𝑺𝑨 + 𝑭𝑨 + 𝑹𝑨)⌉ where 𝑬represents expense categorization accuracy, 𝑷denotes predictive accuracy, 𝑩indicates budgeting efficiency, 𝑺𝑨is savings analysis, 𝑭𝑨is financial awareness, and 𝑹𝑨is risk assessment. Experimental evaluation conducted on 100+ user datasets demonstrates a strong correlation (𝒓 = 𝟎. 𝟗𝟏) with manual financial assessments and an average improvement of 27.5% in user savings behavior after applying AI-generated recommendations. The system maintains secure data handling, real-time insights, and scalability. The proposed platform democratizes intelligent financial planning tools, making them accessible to a wider population.

Web-Based Health Score and Health Calculator System

Authors: Shewale Isha Avinash, Vinod Patidar

Abstract: Remaining healthy becomes increasingly difficult for people due to the changing lifestyle and health- related threats they face. Mostly, all these individuals fail to determine their health condition until they are ill. Health Score is an online solution that analyses a patient’s body mass index, lifestyle, and health records to calculate their overall health level. Information will be gathered from users through the framework’s easy-to-use interface and later processed using the defined health scoring logic. The initial phase involves analyzing the data, followed by normalization to determine the user’s first score. The second phase will involve the use of predictive techniques to classify the users according to their health conditions.

DOI: https://doi.org/10.5281/zenodo.19684377

Crop Yield Prediction And Smart Agricultural Advisory System

Authors: Siddharth Nagesh Gaikwad, Sanket Sadashiv Adling, Danesh Balkrishn Sutar, Mrs. A. G. Chendke

Abstract: Agriculture plays a vital role in the economy of India, acting as the primary source of livelihood for nearly 58% of the country’s population. However, the sector faces significant challenges, including unpredictable weather conditions, pest attacks, and improper fertilizer usage, which often lead to reduced crop yields and economic instability for farmers. This paper presents a comprehensive Crop Yield Prediction and Agricultural Advisory System utilizing advanced machine learning techniques. The proposed system predicts crop yield based on critical environmental factors such as rainfall, temperature, and pesticide usage. Additionally, it integrates specialized modules for disease risk assessment, crop recommendation, and fertilizer recommendation, creating a holistic decision-support platform. The models are trained on large-scale historical datasets. Specifically, the study employs Random Forest for yield and disease prediction, Multinomial Logistic Regression for crop recommendation, and a rule-based system for fertilizer dosage. The system is implemented using the Flask web framework, providing an interactive and user-friendly interface for stakeholders. The experimental results demonstrate that the system can assist farmers in making data-driven decisions, optimizing resource allocation, and significantly improving agricultural productivity and sustainability.

DOI: https://doi.org/10.5281/zenodo.19685934

Iot Based Wearable Monitoring System For Alzheimer’s Patients

Authors: Dr S. Sutha, M. Gowsalya, B. Karthiga, C. Kohila

Abstract: This project focuses on the Alzheimer’s disease is one of the most prevalent neurodegenerative disorders, affecting mil-lions of individuals worldwide and imposing a significant burden on caregivers and healthcare systems. Patients suffering from Alzheimer’s often experience cognitive decline, memory loss, disorientation, and impaired motor function, making them vulnerable to hazards such as falls, wandering, and sudden health deterioration. Traditional monitoring methods rely heavily on manual observation and closed-circuit television (CCTV) systems, which are limited in scope, lack automation, and cannot provide real-time health data.

Smart Farming with Sensors and Automated Pest Control

Authors: Usha S, Kanishkar S K, Nishanth K, Ramakrishnan R, Sakthi S

Abstract: The integration of sensor technologies and automated pest control systems is revolutionizing modern agriculture, ushering in a new era of smart farming. This approach leverages Internet of Things (IoT) devices, such as soil moisture sensors, temperature and humidity monitors, and multispectral imaging, to collect real-time data on crop health and environmental conditions. These insights enable precision agriculture practices, optimizing irrigation, fertilization, and pest management strategies.

DOI: https://doi.org/10.5281/zenodo.19686273

Idea2Team: A Smart Web Platform For Freelancer Hiring And Team Building

Authors: Patel Meet, Patel Smit

Abstract: Finding the right team members and managing collaboration efficiently are major challenges for founders and freelancers. Many users depend on multiple platforms for recruitment, communication, task management, and file sharing, which creates inefficiency. Idea2Team is developed as an integrated web-based platform that simplifies team formation and project collaboration. The platform allows founders to create projects, define required skills, and find suitable candidates through a smart matching system. Freelancers can create profiles, explore projects, and apply based on their skills and experience. Once a team is formed, the system automatically creates a workspace where members can collaborate using real-time chat, task management, and file sharing. The system is developed using React.js, Node.js, Express.js, and MySQL. Real-time communication is enabled using Socket.io. Overall, Idea2Team provides an efficient solution for startup team building and collaborative project execution.

Smart Electric Vehicle Charging Station Locator With Route Planning & Slot Booking

Authors: Gudala Priyanka Sai ramya, Bontha Tarun Kumar, Akumuri George, Mr. A Kabir Das

Abstract: Electric Vehicle Charging Station Locator with Route Planning and Slot Booking using Blockchain is an online platform designed to improve the access, transparency, and reliability of the electric vehicles (EV) charging network. The customers can search the charging stations in a specified city or place, check the slot’s real-time availability, plan the most efficient route, and even book a slot in advance securely to avoid waiting. The blockchain technology ensures the integrity and transparency of booking records as it logs all transactions through cryptographic hashing, thereby creating unalterable, permanent records. This decentralized approach is a win-win situation for both parties, users and administrators, as it not only gives them access to secure and verifiable documentary evidence of booking histories but also helps them to build trust. The admin module includes a centralized dashboard, thereby providing the city, location, station, slot, and user account manager the ability to supervise and control everything, thus, securing data handling and ensuring accurate slot utilization. The system comprises a user interface built in ReactJS, a server coded in Spring Boot (Java), and MySQL for database management. So in conclusion, the system not only provides higher convenience for EV charging but also reduces the waiting period and supports the green electric transport initiative.

DOI: https://doi.org/10.5281/zenodo.19689326

Hire Hub Job Portal System: A Web-Based Recruitment Platform

Authors: Ritik Patel, Raj patel, Dr.Pooja Sapra , Jagruti Parmar

Abstract: The increasing demand for efficient recruitment systems, combined with the limitations of traditional hiring processes, highlights the need for a digital transformation in the recruitment industry. Conventional hiring methods such as manual resume screening, physical interviews, and offline job postings suffer from inefficiency, lack of transparency, and delayed communication. The Hire Hub Job Portal System introduces a modern web-based solution that connects job seekers and employers through a unified platform. The system provides features such as job search, automated application tracking, resume management, real-time communication, and intelligent job recommendations. The platform ensures improved accessibility, faster hiring processes, and enhanced user experience. By integrating modern web technologies, the system creates a scalable, secure, and efficient recruitment ecosystem for both candidates and organizations.

IntelliResume AI Power Screening System

Authors: Posipoyina Jaya Sree, Guna Tirumala Satish, Naimisa Bamala, Gokarakonda Leela Bhavani

Abstract: Organizations now find the hiring process to be inef- fective and time-consuming due to the sharp increase in job appli- cations. Conventional resume screening techniques mostly depend on human labor, keyword matching, and rudimentary Applicant Tracking Systems (ATS), which frequently fall short of capturing individuals’ actual potential and may introduce prejudice. This project suggests an Intelligent Resume AI-Powered Screening System that uses cutting-edge methods from Natural Language Processing and Artificial Intelligence to improve and automate the applicant screening process in order to address these issues. The suggested system uses NLP-based parsing algorithms to extract, evaluate, and interpret data from resumes. It finds important characteristics and transforms them into structured data, including abilities, education, experience, certificates, and projects. The necessary competences are extracted from job descriptions concurrently. The system uses vector representations and semantic analysis to compare resumes with job criteria using embedding techniques and similarity metrics like cosine similar- ity. In order to rank applicants according to their suitability for the position, the system also incorporates machine learning algorithms. Deep learning and transformer-based architectures are examples of advanced models that may be used to enhance contextual comprehension and lessen reliance on precise keyword matches. Additionally, the system has capabilities like suggestion creation, candidate classification (e.g., extremely appropriate, somewhat suitable, and not suitable), and automated scoring. By anonymizing sensitive information like name, gender, and location throughout the screening process, this initiative aims to reduce bias in hiring. This encourages impartial and equitable hiring practices. The system is also scalable and effective at managing high resume quantities, which makes it appropriate for real-world hiring situations. The suggested approach seeks to greatly shorten the time needed for recruiting, increase ap- plicant selection accuracy, and boost overall hiring effectiveness. Additionally, it lays the groundwork for upcoming improvements including integration with chatbot-based hiring platforms, real- time analytics, and adaptive learning models. Thus, in the age of digital transformation, the Intelligent Resume AI-Powered Screening System is a contemporary, effective, and equitable method of hiring.

Review Of Architecture, Challenges, And Prospective Directions For 5G Technology

Authors: Tanisha Jethwa, Smit Chauhan, Param Mehta, Aryan Ramesh Chandran, Dr. Asha Durafe

Abstract: 5G represents the most recent mobile communica- tion technology which aims to improve speed, reduce latency, and offer more reliable connections than past versions. This document provides an insight into 5G network architecture and describes the main network components such as the core network and radio access network. Additionally, the document covers the technologies used in the implementation of 5G communications and different uses cases such as IoT, smart cities, self-driving cars, and telemedicine. The document demonstrates the contribution of 5G technology towards modern digital services. In addition to the benefits of increased throughput, latency reduction, and device density capabilities, several significant problems remain unsolved – from the issue of mmWave propagation restrictions and the energy consumption associated with the densification of networks to compatibility with legacy systems and fragmentation in global standardization. The comparative study demonstrates the advantage of AI/ML-based solutions for autonomous network management. Finally, the document outlines critical research questions and suggests prospective directions for further re- search.

Career Catalyst Ai-Based Online Job Portal

Authors: M.P.P.S.Sai Durga, G.J.Anjana Santhi, K.John Mark, M.Bharath, Shaik M Unnisha Begum

Abstract: Job portal system based on the ideas of Blockchain and Gemini AI to revolutionize the recruitment and employment search process. The program is programmed and consists of three major modules ( Admin, Company and User (Candidate) and is based on the MERN stack (MongoDB, Express.js, React, Node.js).Admin module offers the simplified management of the users, companies, and mock interviewing time, which allows managing the platform efficiently. The Company module enables job advertisements, media of application and shortlisting of candidates, which creates an easy and smooth user experience to employers? The system will offer the candidate (User) a powerful platform, in which he or she could post resumes are subjected to ATS (Applicant Tracking System). It also provides job suggestions on the analysis of skill gap that helps the candidates to match their profiles in accordance with the market demands. One of the most innovative features of the system is the combination of Gemini AI that performs sophisticated analysis in the context of mock interviews, deserves to be mentioned. In addition, the information of the user (and interview) (including resumes and feedback) is securely stored on the Blockchain, connotes integrity, privacy, and transparency of data. Such an incorporation of Blockchain technology ensures that the information of users cannot be tampered and kept confidential.

A Study of Inventory Control Techniques to Optimise Revenue in Small Retail Stores

Authors: Dr Ravendra Kumar

Abstract: According to an estimation, there were around 12.65 million retail grocery stores in India in 2020. These Stores typically covered 100-200 square feet and kept stock of around 500-1000 items. These small retail shops generate yearly average revenue of 18-25 lac, of this the profit margins are around 3-4% i.e., Rs 54,000-72,000 yearly. From discussions with shop owners running small to medium retail stores from two-three generations, the challenge seems to be either overstocking leading to wastage, or stock-out leading to customer diversion. Small retail stores usually are managed using the primitive hit and trial method when it comes to inventory management, that only becomes moderately efficient in case of extensive experience of the manager. The following paper presents a study of three inventory control techniques in order to optimize revenue by controlling the inventory efficiently for small retail stores in India. The purpose of this study is to figure out a pattern connecting inventory control with optimal revenue which may further be converted into an algorithm for software that may help small retail store owners all over India.

DOI: https://doi.org/10.5281/zenodo.19691144

A Machine Learning-Based Framework for Network Traffic Anomaly Detection Using Isolation Forest

Authors: Arjun Raju Polenwar, M. Adarsh Ram, Kembasaram Harini, Dr. G. Venkanna

Abstract: The exponential growth in network traffic volume, velocity, and heterogeneity has rendered manual threat monitoring operationally infeasible for modern enterprise, cloud, and hybrid infrastructures. Conventional signature-based intrusion detection systems (IDS) perform reliably for catalogued attack fingerprints but exhibit critical detection gaps when confronted with polymorphic, obfuscated, or zero-day attack behaviors that have no prior signature representation. This paper presents an anomaly-driven intrusion detection framework grounded in Isolation Forest — an unsupervised machine learning algorithm that isolates statistically rare patterns in high-dimensional data without requiring labeled attack samples during training. The proposed system integrates data ingestion, feature preprocessing, model training, anomaly scoring, quantitative evaluation, and dashboard-based visual reporting into a single reproducible end-to-end pipeline. Validation on an NSL-KDD based benchmark dataset comprising 22,543 records and 42 attributes yields Accuracy = 0.730, Precision = 0.799, Recall = 0.702, and F1-score = 0.747 under label-aware evaluation. These results confirm that a computationally lightweight unsupervised model can serve as an effective first-stage network threat detector while preserving operational interpretability and deployment feasibility.

Design of an Induction Motor for Electric Two-Wheeler Applications

Authors: N. Santosh Kumar, Rohith Patel, CH. Rahul Karthik, R. Nithai Charan

Abstract: Electric two-wheelers are becoming popular due to increasing demand for eco-friendly transportation. Most electric scooters use BLDC motors, but they depend on permanent magnets which increase cost. In this paper, a squirrel-cage induction motor is designed for electric two-wheeler use. The motor is modeled using Altair Flux Motor software. Basic parameters such as dimensions, torque, and efficiency are studied. The motor produces around 3 kW power with efficiency close to 90%. The results show that induction motors can be used as an alternative to BLDC motors for electric vehicles

DOI: http://doi.org/

Design And Evaluation Of An AI Based Adaptive Mock Interview System Using NLP And Real-Time Feedback Analysis

Authors: Himanshu Kumar, Rohit Kumar, Aditya, Prof. Shivangi Patel, Prof. Nitin Pal

Abstract: Preparing for interviews is often inconsistent, as many candidates depend on repeated question lists and general advice that does not clearly show how to improve. This work presents an adaptive mock interview system that creates a more practical way to prepare by combining interaction, evaluation, and guidance in one place. The system is designed to behave like an interviewer by asking questions, examining answers, and giving feedback during the same session. To achieve this, the system processes user responses using language understanding methods and supports both typed and spoken input. Each answer is reviewed from multiple perspectives, including how well it addresses the question, how clearly it is expressed, and the overall tone of the response. Based on these observations, the system assigns a score and provides suggestions that users can apply immediately. A key feature of the system is its ability to adjust question difficulty during the session. When a user performs well, the system gradually increases the level of challenge, while weaker performance leads to simpler or more guided questions. This adjustment helps maintain balance and keeps the user engaged without making the session too easy or too difficult. The system was tested under controlled conditions to observe its behaviour across different types of responses. The results show stable performance, with timely feedback and consistent evaluation. Repeated interaction also leads to noticeable improvement, indicating that the system supports gradual learning and skill development. Overall, the proposed approach offers a structured and flexible way to practice interviews, reducing dependence on manual guidance while helping users build confidence through continuous feedback and adaptation.

A Review Of IoT-Based Smart Home Automation Systems: Insights From Recent Indian Research

Authors: Dipak Sanjay Rothe

Abstract: The accelerating convergence of wireless communication technologies and miniaturized sensor hardware has substantially broadened the scope of intelligent residential systems. This paper presents a systematic review of five peer-reviewed Indian research publications, each examining a distinct dimension of Internet of Things (IoT)-enabled home automation. The studies collectively address control architectures, embedded hardware selection, energy management strategies, real-time hazard detection, and multi-modal user interaction mechanisms. Through structured comparative analysis, recurring themes, prevailing technical constraints, and prospective research trajectories are identified. The review affirms that IoT- driven smart home systems have demonstrated consistent utility in advancing residential convenience, curtailing energy expenditure, and facilitating remote appliance governance, while also highlighting that cybersecurity robustness and large-scale deployment scalability remain underexplored frontiers requiring sustained investigative attention.

DOI: https://doi.org/10.5281/zenodo.19693627

Tradesphere : Multi-Dimensional Trust Platform

Authors: Jashn Patel, 2Bhavesh Parmar, Jugal Khambhata, Raj Howladar, Dr Ritesh Singh Malik

Abstract: Due to the fast-paced emergence of online e-commerce platforms, the landscape of the virtual marketplace has been completely revolutionized by facilitating efficient communication between the buyers and sellers. However, the current systems suffer from a number of trust-related issues such as fake reviews, bias in product ratings, lack of seller responsibility, and lack of means to assess buyer behavior. Rating systems employed in the literature are mainly concerned about the ratings provided by the customers, which leads to an incomplete and often times inaccurate assessment of transaction-based trust. In this research, a Multi-Dimensional Trust Based Rating System (MDTRS) is proposed, where product rating, seller reputation, and buyer credibility are taken into account together in order to form a complete rating mechanism. The proposed rating mechanism differentiates itself from other rating algorithms since it evaluates all the parties involved in the transaction system and hence ensures complete responsibility. In the proposed system, a trust score calculator is developed using dynamic weights according to the credibility of the input and interaction history between the user and the service provider. The implementation of the system is performed through a scalable full-stack architecture along with buyer, seller, and administrator interfaces.

Meachine Learning Approach For House Price Prediction

Authors: Dr. S. W. Mohod, Ms. Falguni N. Mawale

Abstract: House Price Prediction focuses on the development of methods that use machine learning algorithms to accurately predict house prices. Random Forest and Gradient Boosting algorithms have lower mean square error (MSE) and are chosen as the best algorithms for predicting house price. Random forest algorithms handle relationships and provide reliable predictions. Gradient boosting algorithm is used to process large amounts of data to make accurate predictions. Ensemble combines all these individual predictions to produce a final and more accurate prediction. The house information in the dataset also helps improve the estimated house price. This system will help people in the real estate market to make more informed decisions when buying or selling a house.

Comparative Evaluation Of FP32 And INT8 Quantization For Edge Device Real-Time Indian Sign Language Recognition

Authors: Siddharth Roy, Ravish Kumar

Abstract: The high memory and computational requirements of standard floating-point architectures make it difficult to deploy Deep Learning models on edge devices with limited resources. The Post-Training Quantization (PTQ) methods used with a MobileNetV2 architecture for real-time Indian Sign Language (ISL) recognition are compared in this paper. We show that the baseline 32-bit floating-point (FP32) model can be converted to an 8-bit integer (INT8) format, which significantly reduces the model footprint and improves inference latency without significantly degrading accuracy. The INT8 quantized model achieves a 72.4% reduction in memory size (from 9.52 MB to 2.63 MB) and a 31.7% increase in inference speed (from 14.40 ms to 9.83 ms per frame), according to experimental results on a 27-class ISL dataset. Importantly, compared to the FP32 baseline, the quantized model maintains a strong validation accuracy of 98.60%, with an accuracy decline of less than 1%. These results confirm that INT8 quantization is effective in enabling offline, high-framerate computer vision applications on low-power edge hardware.

Federated Time-Series Learning For Cross-Platform Rug Pull Detection

Authors: Dr. Pankaj Malik, Mohit Kapoor, Akshat Gupta, Aman Singhai, Akul Laad

Abstract: The rapid expansion of decentralized finance (DeFi) platforms has been accompanied by a surge in rug pull scams, where malicious actors exploit liquidity pools and abandon projects, causing substantial investor losses. Existing detection approaches are largely centralized and platform-specific, limiting their effectiveness due to privacy constraints, fragmented data sources, and the dynamic behavior of blockchain ecosystems. This paper proposes a novel Federated Time-Series Learning (FTSL) framework for cross-platform rug pull detection that enables collaborative model training without sharing raw transaction data. The proposed system integrates federated learning with advanced time-series modeling to capture temporal patterns in token price volatility, liquidity changes, transaction frequency, and smart contract activities. A hybrid deep learning architecture combining Long Short-Term Memory (LSTM) networks with an attention mechanism is employed to effectively learn sequential dependencies and identify early indicators of fraudulent behavior. The federated setup ensures privacy preservation while enabling knowledge sharing across multiple decentralized platforms. Experimental results on multi-chain DeFi datasets demonstrate that the proposed FTSL model achieves 96.3% detection accuracy, outperforming traditional centralized models (91.2%) and single-platform approaches (88.7%). The model also improves precision (95.1%), recall (94.6%), and F1-score (94.8%), indicating robust and balanced performance. Furthermore, the system is capable of detecting rug pull events 6–12 hours earlier than baseline methods, providing critical early warning signals. Communication overhead is reduced by approximately 28% through optimized federated aggregation, while maintaining scalability across heterogeneous platforms. These findings highlight that Federated Time-Series Learning offers a scalable, privacy-preserving, and highly effective solution for real-time rug pull detection, contributing to enhanced security, transparency, and trust in decentralized financial ecosystems.

DOI: https://doi.org/10.5281/zenodo.19694831

Edge-Aided Acoustic Analysis For Early Detection Of Stem-Borer Infestations: A Multidisciplinary Approach

Authors: Kabisree P, Sandhiya T, Kamalini K

Abstract: Much of the damage caused by stem-boring larvae, specifically the “red palm weevil (Rhynchophorus ferrugineus)”, could be mitigated through “early detection and treatment” of infestations. “Acoustic technology” has the potential to enable this early detection by identifying the short, high-frequency sound impulses produced by larvae as they feed and move within palm tree trunks. However, distinguishing these signals from background noise and wind-induced tapping remains a significant challenge. This paper explores a multidisciplinary approach combining “entomological behavioural analysis”, “advanced signal processing”, and “edge-aided automated machine learning” to provide reliable detection in agricultural environments. By processing signals at the source, these systems can overcome traditional barriers like instrumentation costs and training needs.

DOI: https://doi.org/10.5281/zenodo.19695068

Self-Healing AIoT Systems Using Autonomous Fault Detection And Recovery Mechanisms

Authors: Nafisa S, Dr. Balaji. K

Abstract: The intersection of artificial intelligence (AI) and Internet of Things (IoT) has led to the emergence of massive distributed systems prone to faults due to hardware, network, or software issues. In this paper, we design a self-healing system for AIoT systems that employs automated fault diagnosis, root cause identification, and dynamic healing mechanisms. Our self-healing system uses an efficient transformer-based anomaly detection model coupled with graph neural networks for fault localization and reinforcement learning (RL) agents for recovery policy decision making. The performance of our self-healing framework is evaluated using simulations of a smart factory with 1,200 sensor and actuator nodes, where the framework achieves an average accuracy of 94.2% and latency of 2.3 seconds. When compared to reactive approaches, the framework reduces system downtime by 76.4%, increases task success rate by 15.8 percentage points, and outperforms adaptive recovery approaches by a wide margin.

DOI: https://doi.org/10.5281/zenodo.19697807

Multi-Agent Reinforcement Learning In AIoT For Dynamic Resource Allocation And Optimization

Authors: Nafisa S, Dr. Balaji. K, Shruthi N

Abstract: Due to the rapid rise in the use of artificial intelligence in things (AIoT), the dynamic resource allocation problem is now more complex than ever. In this research paper, a new dynamic resource allocation framework based on multi-agent reinforcement learning (MARL) for AIoT systems is described. A CTDE architecture based on improved MAPPO (IMAPPO) is utilized, which optimizes AoI for action spaces with both discrete and continuous variables. In simulation tests, the framework has achieved 99% successful task completion with 18 MEC servers available, resulting in a minimal probability of failure of 0.01% at 30 dBm. Furthermore, comparative studies show that MARL has better performance than conventional deep Q network (DQN) and proximal policy optimization (PPO) algorithms by 22.01% and 8.26%, respectively. Moreover, a maximum cumulative reward value of 62,306.58 and 99.98% accuracy were obtained, marking a 6.9% improvement over other MARL models.

DOI: https://doi.org/10.5281/zenodo.19698681

Mathematics in the 21st Century: A Review of Key Developments

Authors: Assistant Professor Chavan Akshata Jaydeep, Assistant Professor Dr. G. V. Khandekar

Abstract: Mathematics education in the 21st century is experiencing significant transformation due to rapid technological advancements and evolving pedagogical practices. This review paper examines key developments in mathematics education, focusing on the shift from traditional teaching methods to student-centered approaches that promote critical thinking, creativity, collaboration, and problem-solving skills. The study highlights the growing role of STEM integration and the use of digital technologies, including artificial intelligence, gamification, and online learning platforms, in enhancing student engagement and personalized learning experiences. Furthermore, it discusses major challenges such as teacher preparedness, unequal access to technological resources, and the limitations of traditional assessment systems in evaluating higher-order thinking skills. The review also emphasizes the need for continuous professional development and inclusive educational strategies to ensure effective learning outcomes.

DOI: http://doi.org/

Smart Expense Management Using AI: “Financial Storytelling with Google Sheets Integration

Authors: Yatin Yadav, Prashant Pal, Prince Kumar, Amit Kumar, Mr. Abhishek Chaudhary

Abstract: Personal financial management remains a critical challenge for individuals seeking to track expenses, analyze spending patterns, and achieve savings goals. This paper presents the design, implementation, and evaluation of an AI-powered expense tracking system that combines a user-friendly web interface with intelligent financial insights. The system leverages Streamlit for frontend visualization, Google Sheets as a cloud- based backend, and integrates AI-driven storytelling to provide users with actionable spending analyses. Unlike SMS-based tracking solutions, this system prioritizes privacy, supports cash transactions, and enables rich categorical data entry. Experimental results demonstrate that the system effectively visualizes weekly, monthly, and yearly financial trends while generating meaningful health scores and savings recommendations. The proposed solution offers a secure, scalable, and insightful approach to personal expense management.

DOI: https://zenodo.org/records/19700451

AI Tutor: Personalized Learning Using Visual AI

Authors: Avinash D R, Bharath B R, Deekshith S, Vijay M R, Prof. Madhunandana H M, Dr. Girish Rao Salanke N. S

Abstract: This research introduces a multimodal AI-powered tutoring system designed to make learning more personalized, engaging, and accessible. Unlike traditional digital platforms that provide generic responses, the proposed tutor adapts to each student by understanding their study material and learning needs. It uses low-latency Large Language Models combined with Retrieval-Augmented Generation (RAG) to answer questions directly from uploaded PDFs, ensuring clarity and accuracy. Visual-AI modules interpret diagrams and images to support visual learning, while an automated assessment engine generates MCQs for instant performance feedback. Results show that this approach boosts student confidence, comprehension, and self-directed learning, helping reduce educational gaps and making high-quality tutoring available to everyone.

Lawbot: A Multilingual Legal Assistance Chatbot

Authors: Gayathri Kodipaka, Kompalli Sri Divya Muktha, K. Sandeep

Abstract: Access to legal information remains a significant barrier for millionAs bofstcritaizcetns, especially those who are unfamiliar with formal legal processes or do not speak English. This project proposes LawBot, a multilingual legal assistance chatbot that leverages Natural Language Processing (NLP), a structured knowledge base, and speech recognition to provide accessible legal guidance in seven Indian languages: English, Hindi, Telugu, Bengali, Marathi, Tamil, Kannada, and Malayalam. The system processes voice or text queries from users and responds with relevant legal information drawn from a Firestore-backed offline-capable knowledge base. Upon detecting a legal query, LawBot retrieves a context-sensitive response and delivers it via both text and text-to-speech. This solution is designed for accessibility on both web and mobile platforms and enhances civic awareness by providing a real-time, automated legal information system. Furthermore, it not only enhances user awareness of their rights but also contributes to more equitable access to justice by proactively delivering information in the user’s preferred language. By guiding citizens before complex legal situations escalate, the system can help reduce legal helplessness and promote informed civic participation. Keywords: Natural Language Processing, Multilingual Chatbot, Legal AI, Speech Recognition, Firestore, Knowledge Base, Web Application.

On Road Emission Analyzer With Certification System

Authors: Mrs.Samundeeshwari A, Mr.Kayakakula Saikrishna, Mr.Lingeshwaran V, Mr. Mohan S, Mr.Samraj Sami Durai

Abstract: The project on-road emission analyzer designed to measure real-time exhaust pollutants from moving vehicles using high-precision gas sensors. The system captures key gases such as CO, CO₂, HC, and NOx with improved accuracy through an embedded microcontroller-based data acquisition unit.A wireless communication module transfers the collected data to a central monitoring platform for further analysis. The device enables quick, non-intrusive testing of vehicles under real driving conditions, unlike traditional stationary emission tests. Advanced algorithms validate the measured values against pollution control norms to determine compliance. If the vehicle meets the standards, the system automatically generates a digital emission certificate.

DOI: https://doi.org/10.5281/zenodo.19704959

A Predictive And Simulation Framework For Global Macroeconomics: A Comparative Analysis Of Tree-Based Ensembles And Time-Series Models

Authors: Jay Gavali, Omkar Ghuge, Pratik Kasar, Yash Patil, Jayshree khairnar

Abstract: This paper presents a GDP growth simulation platform, Vikalp.ai, designed for macroeconomic forecasting using machine learning techniques. Traditional economic forecasting methods often rely on linear statistical models, which may struggle to capture the -linear volatilities of global markets, typically relying on rigid linear regressions that falter during sudden economic anomalies. To address this critical gap, this paper details the development and implementation of a robust Random Forest Regression architecture capable of processing over 50 years of historical economic data across 203 countries. By analyzing concurrent indicators such as population dynamics, export and import growth, and capital investment, the study demonstrates how ensemble learning techniques can achieve an exceptional predictive accuracy of ~89.33%. Furthermore, the paper explores the system’s practical application as a real-time scenario simulator, empowering policymakers, researchers, and economists to input hypothetical variables and receive high-confidence GDP growth projections instantaneously. Ultimately, this research bridges the disciplines of artificial intelligence, data science, and macroeconomics, offering a scalable, full-stack web solution to predicting the trajectory of global economies with unprecedented precision, data security, and computational speed.

Holistic vs. Atomistic: A Comprehensive Comparative Study of Generative AI Architectures for Real-Time Personalized Nutrition Coaching

Authors: Vishal Singh, Ajay Rawat, Hemant Singh, Shivam Kumar Jha, Mr. Akhtar Warsi

Abstract: Personalized nutrition coaching applications aim to provide context-aware dietary recommendations aligned with an individual’s health profile, medical constraints, and consump- tion habits. Traditionally, most nutrition analysis engines have relied on modular, rule-based, or ingredient-level classification architectures that process each component of the food label in isolation. With the rise of large language models (LLMs), it has become possible to construct “holistic” architectures capable of processing the entire contextual dataset—ingredients, nutrition table, and user profile—in a single inference call. This paper presents a comprehensive comparative study between two distinct design philosophies: (1) an atomistic Cache-and-Curate architecture based on decomposed multi-step classification, rule chaining, and database lookups; and (2) a holistic Single-Call Generative Architecture inspired by modern LLM capabilities. Using a gold-standard dataset annotated by registered dietitians, we evaluate accuracy, personalization depth, semantic coher- ence, latency, and operational cost. Results demonstrate that the holistic architecture significantly outperforms the atomistic pipeline across all metrics, achieving higher expert alignment, faster inference, richer user-specific reasoning, and substantially lower engineering overhead. The study provides one of the first empirical analyses of architectural paradigms for generative-AI- powered health applications and offers insights into the future direction of real-time contextual reasoning systems.

DOI: https://zenodo.org/records/19706096

User Engagement Analysis In Mobile Learning Applications Using Predictive Analytics

Authors: Lal Rajive Pratap Singh

Abstract: Mobile learning applications are really popular these days for people who want to prepare for exams. One big problem is that people do not stay engaged for long. Most platforms only look at what people did in the past. They do not try to guess what people will do in the future. This paper talks about a way to analyze and predict how engaged people will be with mobile learning applications. We collect data on how people interact with the application like how they use it, how many quizzes they try, and how often they log in. Then we use computer programs, like Logistic Regression, Decision Tree, and Random Forest, to guess if someone will stay active or not. Our results show that the Random Forest program is really good at guessing with an accuracy of 89.3%. This helps us make plans to keep people engaged and improve their learning.

DOI:

Reinventing Standard Costing In The Digital Era: An AI-Driven Empirical Framework For Predictive Cost Management

Authors: Rashna Avinash Golande

Abstract: Standard costing has traditionally served as a key tool for cost control and performance evaluation; however, its relevance has been challenged by the growing complexity of the digital business environment. The emergence of advanced technologies such as artificial intelligence (AI), machine learning (ML), big data analytics, and enterprise resource planning (ERP) systems has significantly transformed conventional costing practices. This study examines the evolving role of standard costing in the digital era, with particular emphasis on current trends in India and global industries. The research adopts an empirical and analytical approach, integrating statistical techniques such as multiple regression and Analysis of Variance (ANOVA) with machine learning models, including Random Forest and Long Short-Term Memory (LSTM). A hybrid AI-driven standard costing framework is proposed and evaluated using cost-related data. Model performance is assessed using error metrics and further validated through an ablation study. The findings indicate that the integration of AI and predictive analytics improves cost estimation accuracy by approximately 25–30 percent compared to traditional methods. The study concludes that standard costing is evolving into a dynamic, real-time, and predictive system, enhancing both operational efficiency and strategic decision-making in the digital economy.

NeuroFit : An AI-Powered Web Application For Body Type Detection And Personalised Fitness Guidance

Authors: Shubham Bawari, Paras Chandra

Abstract: The NeuroFit application before was using an CNN in web browser to decide the body type by looking at webcam pictures, and after that, Google Gemini system was used for producing a fitness plan. Even if this way worked, it always needed internet to be on, so it failed sometimes for students having a bad connection. With version two, the main system changed. Now classification for body type used a Random Forest which looks at seven biomechanical measurements that are coming from the Media Pipe Pose points instead of using original image pixels. Dependencies like Media Pipe WASM files and other necessary files come with the app so after the first setting up there are not any internet calls. Instead of three possible body types, the classifier increased decision categories to four BMI groups based on WHO: Underweight, Normal Overweight and Obese. The algorithm learned patterns from 6,000 made-up examples with classes mixed at BMI cutoffs, and gets 82 percent accuracy on the test samples. Height error for 30 sample testers was an average of 4.2 centimetres. Usability for 25 AKTU students got a SUS mark of 81.2 which was more than first version’s 78.5. The system’s countdown was better too, increasing from 60 to 94 percent after updating thresholds in frame. The main discovery is that the Random Forest with landmarks can be explained with feature importance calculation, is accurate and does not need the network.

Neurfit Ai Fitness Website

Authors: Shubham Bawari, Paras Chandra

Abstract: The healthcare landscape is experiencing a transformation toward individualized wellness solutions. This paper introduces a comprehensive health optimization ecosystem that leverages contemporary machine learning methodologies and artificial intelligence capabilities to formulate adaptive exercise regimens, customized nutritional strategies, and continuous health progress tracking. The framework processes individual attributes such as demographic data, biometric measurements, fitness aspirations, and dietary inclinations to construct tailored health roadmaps. Our technical infrastructure integrates React with Redux for user-facing components, Python-based backend services utilizing Flask and Node.js for backend operations, coupled with MongoDB for persistent data storage. The platform incorporates AI-driven analytical capabilities alongside interactive engagement mechanisms, peer networking functionalities, and granular monitoring tools to sustain user motivation and behavioral change. Validation efforts demonstrate the platform’s effectiveness in delivering context-aware, responsive wellness guidance with potential for significant lifestyle improvement. This investigation explores the intersection of artificial intelligence and personal health management, highlighting opportunities for enhancement across accuracy dimensions and end-user experience quality.

Big Data: Techniques, Challenges, Applications, And Scope

Authors: Mrs. Shwetambari Sukhadeo Dhupe

Abstract: The continuous advancement of digital platforms has resulted in the generation of extremely large and complex datasets, collectively known as Big Data. Conventional data management and processing techniques are no longer adequate to manage such datasets due to their size, speed, and diversity. Big Data analytics employs distributed architectures, scalable storage solutions, and intelligent analytical models to derive valuable insights. This paper presents a concise framework on the fundamental characteristics of Big Data, supporting technologies, its application across various sectors, major scope along with emerging research directions.

DOI:

Integrated Fire Suppression Robot With SmartRainwater Harvesting

Authors: Mr. Nithin AC, Mr. Sharanagouda, Mr. Gurukiran G K, Mr. Bhuvan H M, Dr. Nanda B S

Abstract: This project focuses on developing a smart fire suppression robot integrated with a rainwater harvesting system to improve safety while making better use of available water. The system automatically collects rainwater during rainfall and stores it in a tank, while continuously monitoring the water level for effective usage. The robot is equipped with flame sensors to detect fire and can respond either autonomously or through Bluetooth control. A compact water pump draws water from the tank and sprays it through a movable nozzle, allowing accurate and effective fire suppression. The design is simple, portable, and easy to use, making it suitable for real-world applications. It reduces the need for constant human involvement and helps ensure a quick response during emergencies. In addition, the system is built using low-cost and easily available components, making it affordable and practical for wider use. By combining rainwater collection with automated fire response, the project provides a reliable and efficient solution that supports both safety and better use of natural resources.

Machine Learning – Working Principles, Applications, Advantages, And Techniques: Review

Authors: Mrs. Shobha Chandrakant Kadam

Abstract: The rapid-fire advancement of machine learning( ML) has brought transformative changes across various industries, including healthcare, finance, manufacturing, independent systems, and retail. The exponential growth of data and computational power has enabled the development of largely sophisticated ML algorithms able of making precise prognostications and enhancing decision- making processes. still, the widespread relinquishment of ML also brings forth critical ethical enterprises, similar as data sequestration, algorithmic fairness, and the future of employment in a largely automated world. This paper presents a comprehensive review of ML by examining its working principles, significant applications across diligence, core advantages, and foundational ways. It explains how ML models are trained, validated, estimated, and stationed in real- world scripts. likewise, this paper outlines the primary ML paradigms, similar as supervised, unsupervised, reinforcement, and deep learning, offering insight into their roles in contemporary technological invention.

DOI:

Digital Entrepreneurship and E-Commerce Growth in India

Authors: Atharva Yadav, Hardik Solanki, Teesha Wala, Shaikh Daanish Ahmed, Prof. NandKishor Narkhede

Abstract: Rapid expansion in the use of digital technology has brought significant changes in the business environment in India, leading to the development of digital entrepreneurship. Moreover, the e-commerce industry in India has also witnessed a significant growth due to greater connectivity, higher mobile phone usage, and shifts in consumer behavior. This research paper aims at examining the correlation between digital entrepreneurship and growth in the e- commerce industry in India. In this regard, the analysis will be conducted on the following areas: Drivers of growth in the e-commerce industry. Based on the analysis from relevant literature and other secondary sources, it can be observed that the use of digital platforms has made it easier for micro, small, and medium-sized enterprises (MSMEs) to access the market both nationally and internationally. Examples such as Flipkart, Meesho, and the Jio revolution are used to provide context based on theoretical concepts. Key barriers, including lack of digital infrastructure, cybersecurity issues, and complex regulations, among others, have been noted.

Control Block In Robotic System

Authors: Shivali Sawant, Veeksha Shetty, Gaurang Parmar, Jainish Shah, Nandkishor Narkhede

Abstract: This paper presents a case study on the implemen- tation of Process Control Blocks (PCBs) in robotic systems for efficient task management. Modern robotic systems must simul- taneously manage multiple processes including sensor data col- lection, motor control, and inter-system communication. Without structured process management, resource conflicts can degrade system performance and delay critical tasks. This study analyzes how PCBs enable operating systems to organize, schedule, and switch between processes effectively in robotic environments. Through analytical investigation and review of existing literature, the study demonstrates that PCB-based scheduling significantly improves robot reliability and real-time responsiveness. Schedul- ing algorithms such as Priority Scheduling and Round Robin are evaluated in the context of robotic operations.

HIRKANI: An AI-Powered Food Safety Guide For Pregnant Women

Authors: Atharv Mane, Suyash Thombare, Pratik Patil, Varadraj Nangare, Saniya Attar, Dr. Jagruti Jankar

Abstract: Pregnancy is a vital stage for any woman. The dietary preferences significantly influence the mother’s health and the growth of the baby. Nevertheless, pregnant women in India face difficulties in understanding whether a particular packaged food is safe for their health during pregnancy. The reasons include complicated labeling systems, confusing scientific names of ingredients used in the food products, and absence of specific information regarding the pregnancy-related safety of such products. Thus, this research paper introduces HIRKANI – a web application, which combines AI technologies and other approaches for providing timely and evidence-based information about food safety for pregnant women. Specifically, HIRKANI applies the barcode scanning technique, OCR algorithms, and image recognition performed by AI for identifying different food products and assessing their safety. The product safety rating is evaluated using a rule engine that considers the recommendations issued by FSSAI, ICMR-NIN, and WHO. In total, the application distinguishes between safe, consume with caution, and avoid food products

Development And Quality Characterization Of A Functional Confectionery: Physico-chemical And Sensory Evaluation Of Beetroot (Beta Vulgaris L.) Powder-Incorporated Besan Laddu

Authors: Nazia Hussain, Kh. Pinky Devi, Takhellambam Ranjita Devi, Kshetrimayum Vedmani Devi

Abstract: Background: Beetroot (Beta vulgaris L.) is some nutritionally dense root vegetable rich in betalains, inorganic nitrates, and essential minerals. Despite its therapeutic potential in combating oxidative stress and improving cardiovascular health, its application in traditional Indian confectionery remains limited. Objective: This study aimed to develop a value-added besan laddu incorporated with beetroot powder and to evaluate its sensory, physical, and biochemical properties, along with consumer acceptability and economic feasibility. Methods: Four formulations were developed: a control sample and three experimental variations incorporating beetroot powder at concentrations of 30 g, 40 g, and 50 g. Sensory evaluation was conducted using a 9-point hedonic scale by a semi-trained panel (n=10), and broader consumer acceptability was assessed using the Food Action Rating (FACT) scale (n=100). Statistical significance was determined using the Friedman test (p = 0.406). Results: Variation 2 (100 g besan, 40 g beetroot powder, and 60 g jaggery) emerged as the most acceptable formulation, achieving superior scores for taste (7.8 ± 0.62), flavor (7.6 ± 1.37), and overall acceptability (7.6 ± 0.71). Physical analysis of the optimized product revealed an average weight of 21.3 g, volume of 24.3 ml, and a bulk density of 0.86 g/ml. Biochemical profiling of the selected variation showed a moisture content of 15.69%, fat content of 13.98%, and total ash of 3.74%. Consumer evaluation yielded a FACT score of 7.15, indicating high market potential. Conclusion: The study demonstrates that moderate incorporation (40%) of beetroot powder significantly enhances the functional profile of traditional laddu without compromising sensory integrity. At a production cost of ₹110 per 100 g of raw ingredients, this product represents a viable strategy for delivering plant-based bioactive compounds through traditional snack formats.

DOI: https://doi.org/10.5281/zenodo.19731335

Planning And Cost Accounting Of Construction Equipments

Authors: Ms.Sharvari Raju Gawde, Prof. N.R.Patil

Abstract: Construction projects require efficient planning and cost control to achieve timely completion and economic feasibility. This study focuses on the planning and cost accounting of construction equipment used in large-scale projects, with a detailed case study of the Koyna Hydro Electric Project (Stage IV extension). The research emphasizes the importance of selecting appropriate plant and equipment based on project requirements, site conditions, and economic considerations. The study examines the balance between mechanization and manual labor, highlighting that proper equipment selection significantly influences productivity, cost efficiency, and project duration. It also analyzes ownership and operating costs of equipment, including factors such as depreciation, fuel consumption, maintenance, and utilization rates. A comparison between estimated hourly usage rates and actual costs incurred is carried out to identify variations and improve future decision-making. The findings indicate that effective equipment planning, optimal utilization, and accurate cost accounting can reduce overall project costs and enhance operational efficiency. The study concludes that systematic management of construction equipment plays a vital role in achieving quality work, timely completion, and cost optimization in modern construction projects.

Deep Fake Audio Detection Using MFCC And NLP

Authors: Mrs. M. Devika, UG Student

Abstract: The recent developments in restrictions on AI have resulted in the ability to create almost realistic-sounding deep fake audio content which has created new threats to online safety, security and privacy with many kinds of identity crimes or fraud, misleading or erroneous information, impersonation. The vast majority of audio recordings made by real people sound different than those produced synthetically with subtle variations in voice characteristics; therefore, it is becoming increasingly difficult for traditional methods to accurately identify any audio as being ‘real’ or ‘fake.’ Consequently, the design of an AI-powered method for detecting Deepfake Audio will utilize Mel Frequency Cepstral Coefficients (MFCCs) for feature extraction (from both samples of voice recordings) using Natural Language Processing (NLP) incorporating an analysis at the level of speech to determine whether or not a sound is legitimate based on their similarities and differences to other sounds and how the two audio files relate to one another within the context of their use. Artificial Intelligence will aid in achieving this objective through the extraction and classification of the spectral and linguistic features contained within the audio files by two separate models: Machine Learning (ML) and Deep Learning (DL). The solution has been demonstrated to improve detection rates and to exploit multiple types of media that can be employed for producing deep fake audio attacks; hencely making it extremely resilient against attacks of all types. Therefore, the technology has the potential for augmenting the safety, reliability and confidence level that individuals have when using voice-based digital communications.

TotalShield: A Multi-Layered Defense Architecture For Robust Protection Against Prompt-Leaking Attacks On Large Language Models

Authors: Sanchayan Ghosh

Abstract: Large Language Models (LLMs) are increasingly being used in today’s real world artificially intelligent applica- tions, but they are still vulnerable to prompt leakage attacks which results in intellectual system prompt extraction. PLeak is such an advanced attack framework, where attackers try to obtain the system prompt through the model responses by carefully crafting adversarial queries iteratively. Currently available defense strategies, such as alignment based filtering, subspace constraining and casual isolation, are only partially effective against adaptive attacks, replay attacks and semantically obfuscated attacks. In this paper, we propose TotalShield, a comprehensive multi layered defnse system that aims that aims to provide strong immunity to prompt leakage attacks without solely depending on fine tuning the models. TotalShield is a multi layered defense system that consistes of seven synergistic protection layers: Prompt sanitization, Prompt fingerprinting, Neuro-guard transformation, concept masking, Leakage scoring, Adversarial behavior detection and Post generation rewriting. Unlike traditional single layer defense frameworks the proposed system combines lexical filtering, behavioral heuristics, semantic inspection, and response abstraction within a unified pipeline. To test robustness, we conduct iterative PLeak attacks and measure system performance using four quantitative metrics: Exact Match (EM), Substring Match (SM), Edit Distance (ED), and Semantic Similarity (SS). Our experimental result show that TotalShield preserves zero exact and partial leakage while maximizing semantic distance from protected system content. TotalShield is a deployable security solution for protecting system prompts in LLM services. By integrating deterministic security with semantic risk analysis, TotalShield pushes the frontiers of prompt leakage defense towards fully hardened systems.

DOI: https://doi.org/10.5281/zenodo.19750397

Advanced Url Scanner For Industry 4.0 Applications

Authors: Ms.V.Sailaja, Shashikiran Begari, Shaik Mahaboob Subani Tahameer, Charan Yarlagadda

Abstract: Rogue websites have become one of the most important means of cyberattack, such as phishing, malware distribution, credential theft, and ransomware attacks. Cybersecurity has become a serious issue in the age of Industry 4.0, when corporations become more and more dependent on networked digital platforms and cloud-based services. Conventional URL filtering systems, based on static blacklists and rule-based systems are not able to see zero-day threats and newly created malicious domains. To address these shortcomings, we believe in an Advanced URL Scanner which carries out real-time, multi-layered security scanning of the web URLs. The system combines threat intelligence services, validation of SSL certificate, tracking of redirect chain and analysis of IP geolocation. The system also has AIs that analyze the webpage content to detect phishing schemes and malicious activities. The backend system is built on top of node.js and express and the front-end dashboard is developed on top of react.js to provide real-time analytics view. Experimental test shows that the system has a high detection rate and low processing latency, which makes it appropriate to reinforce cybersecurity systems in Industry 4.0 digital spaces.

DOI: https://doi.org/10.5281/zenodo.19754178

System On Chip Design And Challenges

Authors: Khushi Gupta, Dr. Asha Durafe, Nisha Kambli, Jay Gori, Shashank Jethva

Abstract: The System-on-Chip (SoC) technology has become an important topic in current electronics engineering. The present study highlights various issues faced during the design of an SoC. These include scalability, energy consumption, reliability, and difficulty in integration. An extensive review of the literature has been carried out for methodology identification.

Atom: The Medical Chatbot With Voice Assistance

Authors: Raj Tilak Singh

Abstract: We are thrilled to introduce you to an innovative way of making health care more convenient and interesting with our Medical Assistant Chatbot based on Natural Language Processing. We can convey our health concerns, symptoms, and queries with this intelligent chatbot in an as real time manner. Our Natural Language Processing which is a based chatbot can understand what you are saying and can help you in finding out what illness you might be suffering from, what symptoms you might be experiencing, etc. But that is not all. It also helps you remember when you should be taking your medicines. On top of that, the app also includes a simple map that you can use to find the nearest hospitals that are best suited for your needs depending on your symptoms. The app combines mapping with intelligent suggestions for your health, ensuring that you are able to quickly determine the best course of action if you are in need of care. Rather than trying to search for the information on your own, you are able to get instant advice, making the entire experience much easier for you

DOI: https://doi.org/10.5281/zenodo.19754697

AI-Based Real Time Predicting Employee Attrition

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

Abstract: Employee attrition is a significant issue for organizations across the globe, as it results in increased recruitment costs, diminished skilled employees, and decreased organizational productivity. Early identification of employees that are likely to leave the organization can provide proactive action and enable Human Resource (HR) departments to undertake retention strategies. In this paper, we provide a machine learning-based solution to predict employee attrition, and we focus specifically on the Support Vector Machine (SVM) algorithm as the primary machine learning model to create the predictive model. We also trained several SVM kernels and conducted hyperparameter tuning to enhance the accuracy and capability of generalization. We used multiple performance metrics to evaluate the models such as accuracy, precision, recall, F1-score and ROC-AUC. Our experiments demonstrated that the SVM predictive model consistently outperformed the performance of other models by accurately identifying the cases of higher risk of attrition, allowing us to conclude that it was a sufficiently reliable model. To ensure ease of use and accessibility we developed a web application utilizing Streamlit, while maintaining a user-friendly interface. The application allows human resource professionals to enter employee characteristics and receive real-time predictions of potential employee attrition and benefits HR professionals by comparison of predictive models and performance measures used, as well as visualizations of performance metrics and various datasets. This paper aims to connect the gap between data science and HR decision making. This paper applies powerful machine learning methods to an interactive software interface and produces a straightforward, scaleable yet intelligent system to assist organizations with employee retention, workforce planning, and long-term strategic growth.

AI-Powered Sign Language Converter Using Image Recognition Techniques For Smart Home Control

Authors: Ameya Katkar, Riya Suryavanshi, Harshad Kadam, Purva Kamat, Prof. M. S. Chavan

Abstract: The demand for assistive technologies and smart home automation had increased the necessity of developing a system with effective and real-time human- computer interaction. This paper proposes an AI-based sign language translator for smart home automation, where the user can use hand gestures to control home appliances. We used a computer vision-based, deep learning IoT embedded system to design a simple yet effective system for accurate and low-latency gestural identifications. Unlike raw image-based approaches, the system uses vision-based input acquired by a camera, from which MediaPipe extracts 21 hand landmarks for every frame. For identifying dynamic gestures, one approach is to use a hybrid CNN–LSTM model that learns spatial and temporal features from sequences of gesture data. Trained on gesture sequences of 30 frames per sample, the model obtains an average recognition accuracy ~95% on the test dataset. The system operates on an edge device which runs on Raspberry Pi technology to deliver its real-time function while achieving cost- effective and energy-saving results. The system operates in real-time environments because it processes each gesture with an average inference time of one to two seconds. The system uses relay modules for gesture recognition to create control commands which are sent to IoT devices to perform actions such as activating buzzers and controlling fans and switching lights on and off. The suggested method achieves precise results through landmark-based processing which decreases computational needs by 70 to 80 percent compared to standard image-based deep learning methods. The system performs well in a variety of backgrounds and lighting situations, making it appropriate for real-world implementation. The proposed method provides a flexible solution which operates with short delays and low costs to create smart home automation systems that assist disabled users to achieve independent living.

Wearable Health Monitoring System: A Multi-Sensor IoT Architecture With Edge Intelligence And Cloud-Based Anomaly Detection

Authors: Neal Bheda, Dr.Asha Durafe, Neeil Mahadik, Heet Thakkar, Jeel Mange, Dhruv Trivedi, Aayushi Bhanushali

Abstract: Wearable health monitoring systems (WHMS) rep-resent a transformative paradigm for continuous, non-invasive physiological surveillance, enabling early detection of life-threatening conditions outside traditional clinical settings. This paper presents the design, implementation, and clinical eval-uation of a compact five-modality WHMS that concurrently acquires electrocardiographic (ECG) waveforms, photoplethys-mographic (PPG) oxygen saturation (SpO), infrared skin-surface temperature, barometric-pressure-derived blood pressure esti-mates, and six-axis inertial motion data. An on-board Raspberry Pi Zero 2W edge processor executes real-time QRS detection, four-class LSTM arrhythmia classification, adaptive-filter SpO computation, and random-forest fall detection before forwarding compressed feature vectors via Bluetooth Low Energy 5.0 to a paired smartphone gateway and thence to a HIPAA-compliant cloud backend. Clinical validation across 22 volunteers against gold-standard reference instruments yielded ECG heart-rate MAE of 0.7 bpm, SpO MAE of 1.1%, systolic blood-pressure MAE of 4.8 mmHg (meeting AAMI SP-10), arrhythmia classifi-cation macro-F1 of 95.4%, fall-detection sensitivity of 97.3%, and an operational runtime of 11.2 hours under a realistic mixed-activity duty-cycle profile. These results demonstrate that deeply integrated edge-cloud WHMS architectures are capable of meeting clinical accuracy standards within a consumer-wearable form factor.

Real Time Blood Bank Inventory Management And Request System

Authors: Surya Vallabaneni, Ranjith Reddy, Shivansh Behal, Vashu Satya Prakash

Abstract: Blood availability in emergency is a major obstacle in modern medical care systems. Without real time tracking these traditional systems are prone to the significant in matching supply with demand on that particular time, it effects lack of real time tracking, leads to inefficiencies in matching blood supply, improper communication, and less accessibility for the users. These limitations are often result in the delays that can seriously impact on patient medical care. This research proposes the Real Time Blood Bank Inventory Management and Emergency notification Alert to the System that integrates smart phone applications with a cloud-based data structure to check the blood availability and cooperation during emergency situations. This proposed application system uses the Node.js backend, Firebase Fire store database, and a smart phone base application to enable the real time happen at the exactly same time of blood availability of detailed list across multiple blood banks. It incorporates the advanced features such as multi sections search based on the blood type and location, emergency alert notifications, donor registration page, and the rol based access control for a secure operation. unfortunately, the blood bank application system that improves communication between the hospitals, donors, and blood banks through give control of all the parts of a platform. The act of putting a plan of this medical care system significantly improves the response time, improves transparency, and ensures enough utilization of the available blood resources. Experimental analysis the demonstrates faster access to the blood availability information and better coordination compared to traditional medical care systems. This research gives modern healthcare by giving data base, real time decision making and also providing a large scalable and reliable solutions for blood bank management system.

Agriculture Portal

Authors: Amit Gupta, Tushar Karbhar

Abstract: The Agriculture Portal is an innovative web-based application designed to connect farmers directly with customers, eliminating the need for middlemen and ensuring fair pricing for agricultural produce. The platform provides a structured marketplace where farmers can list their crops, while customers can easily browse and purchase products. With dedicated modules for farmers, customers, and government authorities, the system enhances efficiency and transparency in the agricultural sector. Farmers can manage their profiles, track sales, and access valuable market insights, while customers can directly engage with producers for fresh and high-quality goods. To further support farmers, the portal integrates multiple advanced features such as weather forecasting using the OpenWeatherMap API and a machine learning-powered crop prediction system to help farmers make informed cultivation decisions. A built-in news feed ensures users stay updated with the latest agricultural trends, government schemes, and market rates. Additionally, the platform offers multilingual support, including languages like Marathi, making it accessible to a diverse range of users. The system is secured with two-factor authentication, ensuring data privacy and safe transactions for all users. Technically, the Agriculture Portal is developed using HTML5, CSS, Bootstrap, JavaScript, and jQuery for the frontend, while the backend is powered by PHP and Python, with MySQL handling data storage. Secure and seamless transactions are facilitated through the Stripe payment gateway, while APIs such as SendGrid for email services and News API for real-time updates further enhance functionality. By integrating modern technology with agriculture, this platform serves as a one-stop solution for improving the efficiency, profitability, and sustainability of farming practices.

Smart Helmet for Mining Workers with Real Time Monitoring and Hazard Detection

Authors: Haripriya K, Boopathi G, Tamilarasan V, Harivasan S

Abstract: Hazardous industrial and mining environments pose significant risks to worker safety due to exposure to toxic gases, extreme environmental conditions, and physical stress. This paper presents the design and implementation of an Internet of Things (IoT)-based Smart Mining Helmet aimed at enhancing occupational safety through continuous monitoring and intelligent alert mechanisms. The proposed system integrates multiple sensors, including a gas sensor for detecting harmful gases, a temperature and humidity sensor for environmental monitoring, a vibration sensor for detecting abnormal motion, and a pulse sensor for monitoring the physiological condition of the worker. The system utilizes Wi-Fi connectivity and the Message Queuing Telemetry Transport (MQTT) protocol to transmit real-time sensor data to a cloud-based dashboard, enabling remote supervision. A GPS module is incorporated to provide location tracking, facilitating rapid response during emergency situations. The system performs threshold-based analysis of sensor data, and upon detecting unsafe conditions, it generates immediate alerts and notifications. Furthermore, the collected data can be logged and exported for further analysis, supporting predictive safety measures and decision-making. Experimental evaluation demonstrates that the proposed system reliably monitors environmental and physiological parameters and provides timely alerts under hazardous conditions. The solution is cost-effective, scalable, and suitable for deployment in real-world industrial applications.

DOI: https://doi.org/10.5281/zenodo.19756509

TerraSecure: A Machine Learning Framework For Detecting Infrastructure As Code Misconfigurations With 10.7% False Positive Rate

Authors: Ms.Dhivya K, Bhavayazhinitha S V, Gunal S, Jashwanth M U, Kanishka R, Gokulnath K

Abstract: However, the misconfiguration in the IaC template is now regarded as one of the critical factors responsible for cloud security breaches. Services impacted by these include storage, networking, identity management, and databases. This paper discusses TerraSecure – an advanced intelligent multilayer framework that is capable of identifying such misconfigurations. TerraSecure applies a hybrid approach which includes rule-based detection, machine learning, and AI-powered contextual analysis. This framework employs more than 50 security patterns extracted from actual breaches as well as best practices in cloud computing. A pre-trained XGBoost model, considering 50 security patterns, predicts the risk score with accuracy of 92.45% while ensuring the minimal false-positive ratio of 10.71%. As a result, vulnerable configurations, such as public storage access, overly broad permission scopes, unencrypted data, and unsafe network settings, can be identified. Moreover, an AI analysis component adds to the interpretability of this framework by delivering information about potential business impact, attack scenario (based on the real incident), and remediation steps. In addition, TerraSecure supports several output formats, among which there is SARIF to facilitate the integration with CI/CD pipelines and other tools (e.g., GitHub). The conducted experiments confirm the scalability, efficiency, and reliability of this framework in terms of security.

DOI: https://doi.org/10.5281/zenodo.19756986

Deepfake Voice Detection Using CNN And LSTM

Authors: Baby Saral G, Vibin Dev Anand, Hariharan A, Nehaansh Ladiwala

Abstract: Voice synthesis has come a long way. In just a few years, AI-generated speech has gone from obviously robotic to genuinely convincing—convincing enough that distinguishing a real recording from a synthetic one is no longer something a casual listener can reliably do. That shift creates a real problem for voice authentication systems, digital media verification, and any context where the authenticity of an audio recording actually matters. In this work, we explored whether a hybrid CNN–BiLSTM architecture, operating on Mel spectrogram representations, could reliably separate genuine human speech from AI-generated audio. The idea was to let convolutional layers pick up on spectral irregularities that synthesis systems tend to leave behind, while bidirectional LSTM layers model how those patterns evolve across time—something a frame-by-frame analysis would miss entirely. We trained and evaluated the model using the Fake-or-Real (FoR) dataset, which contains around 17,870 labelled clips split evenly between authentic recordings and outputs from various neural TTS systems. On the held-out test partition, we observed an overall accuracy of 99.02%, with precision, recall, and F1-score each sitting at 0.98, and an AUC of 0.9998. Honestly, the AUC was higher than we anticipated going in. To check whether the recurrent component was actually helping, we also ran a CNN-only version under identical conditions—it dropped to 96.4% accuracy, which suggests the temporal modelling is doing something the convolutional layers alone cannot. We describe the full pipeline here, from raw audio input through to REST API deployment, and we try to be upfront about where this approach currently falls short.

DOI: https://doi.org/10.5281/zenodo.19757403

Natural Language Processing (NLP) In Artificial Intelligence

Authors: Miss.Sandhyarani Prakash Randive

Abstract: Natural Language Processing (NLP) is a technology that allows machines to become more human-like, narrowing the gap between humans and machines. In a nutshell, NLP allows humans to converse with machines more readily. NLP has a wide range of applications that have been developed during the last few decades. The majority of these are really helpful in everyday life, such as a machine that accepts voice commands. Many research organizations are working on this problem in order to produce more practical and usable solutions. Natural Language Processing has a lot of potential for producing computer interfaces that are easier to use for humans, because people will be able to communicate with computers in their own language rather than having to learn a computer language.

Smartphone-Assisted Deep Learning Model For Oral Cancer Screening At Rural PHCs

Authors: Rubina Begam M, Manikandan Sathyamoorthy, Mohamed Akram. M, Mohamed Faizal. B, Magudesh. M

Abstract: Oral cancer is a significant public health concern, especially in rural areas where access to specialized diagnostic facilities is limited. Early detection is crucial for improving survival rates and reducing treatment costs. This paper presents a smartphone-assisted deep learning model for oral cancer screening at Rural Primary Health Centers (PHCs). The proposed system utilizes oral lesion images captured using smartphone cameras and applies deep learning techniques for automated classification of normal and suspicious cancerous lesions. Image preprocessing methods are employed to enhance quality and improve feature extraction, while a Convolutional Neural Network (CNN) model is trained to achieve accurate lesion detection. The integration of smartphone technology with artificial intelligence enables a low-cost, portable, and user-friendly screening solution suitable for resource-limited settings. The developed model assists healthcare workers in performing preliminary screening and supports timely referral for further diagnosis and treatment. Experimental results demonstrate promising accuracy and reliability in detecting oral abnormalities, showing the potential of the proposed system as an effective screening tool. This approach bridges the gap between advanced diagnostic technologies and rural healthcare delivery, contributing to early detection, improved accessibility, and reduction in oral cancer burden.

DOI: https://doi.org/10.5281/zenodo.19759303

E-Pharmacy Ecosystem

Authors: Rajat Nirwal, Tapanshu Dayal Tyagi, Sambhav Jain, Ankur Kaushik

Abstract: India’s rapidly digitizing healthcare ecosystem presents a significant opportunity for scalable, affordable, and regulation-compliant e-pharmacy platforms. However, current implementations largely focus on transactional commerce rather than regulatory engineering, real-time interaction, and intelligent cost optimization. This paper proposes a compliance-first, modular MERN-stack-based e-pharmacy architecture integrating: • AI-assisted prescription digitization • Real-time WebSocket communication • Intelligent generic drug substitution • Pharmacist-supervised validation workflows • Embedded regulatory governance mechanisms The system aligns with statutory frameworks including the Drugs and Cosmetics Act (1940), Telemedicine Practice Guidelines (2020), and the Digital Personal Data Protection Act (2023). Unlike traditional monolithic e-commerce systems, the proposed model introduces a service-oriented MERN architecture with mathematically modeled substitution logic and hybrid verification pipelines. This research contributes a scalable, regulatorily aligned, AI-integrated blueprint for next-generation Indian digital pharmacy ecosystems.

The Django Framework in Python: Architecture, Features, and Applications

Authors: Sandhyarani Prakash Randive

Abstract: The rapid growth of web applications has increased the demand for efficient, scalable, and secure development frameworks. Django, a high-level Python web framework, has emerged as a popular choice due to its simplicity, robustness, and adherence to best practices. This paper explores the Django framework, its architecture, core components, features, advantages, and real-world applications. It also includes architectural diagrams to illustrate Django’s Model– View–Template (MVT) design pattern. The study highlights Django’s role in modern web development and evaluates its strengths and limitations.

A Real-Time Edge-Based Deep Learning System For Automated Dental Cavity Detection And Risk Prediction Using Yolov5 And Raspberry Pi

Authors: Ashish Parekh, Devesh Tomar, Dhruv Selopal, Jenil Patel, Prof. Biju Balakrishnan

Abstract: Early detection of dental caries is essential for preventing severe oral complications, yet conventional diagnostic approaches rely heavily on manual visual examination and radiographic interpretation, which are often subjective and resource-intensive. This paper presents a real-time edge-based deep learning system for automated dental cavity detection and risk prediction using a Raspberry Pi-powered intraoral imaging platform. The proposed framework integrates a high-resolution camera module with a YOLOv5 object detection model trained on annotated dental image datasets. The system performs on-device inference, enabling real-time cavity localisation without reliance on cloud infrastructure. A confidence-based filtering mechanism reduces false positives and improves diagnostic reliability. A lightweight risk prediction module analyses historical detection patterns to assist in preventive dental assessment. Experimental validation demonstrates strong agreement between model predictions and expert annotations, confirming the system’s reliability and feasibility for deployment in resource-constrained environments.

DOI: https://doi.org/10.5281/zenodo.19760227

GUI-Based Weather Dashboard Using Python And OpenWeatherMap API

Authors: Sandeep Maurya, Mrs. Pooja Singh

Abstract: Weather information is one of the most frequently accessed categories of data in everyday life. Despite the widespread availability of weather applications on smartphones and the web, there remains a notable gap in the availability of dedicated, desktop-grade, Python-based weather monitoring tools that integrate real-time data with local analytical capabilities, air quality reporting, and persistent data management. This paper presents the design, architecture, and implementation of a GUI-Based Weather Dashboard developed using the Python programming language, the Tkinter graphical user interface framework, and the OpenWeatherMap API for real-time meteorological and air quality data retrieval. The proposed system provides users with a comprehensive view of current atmospheric conditions, including temperature, humidity, wind speed, atmospheric pressure, and a textual weather description. Additionally, the dashboard incorporates an Air Quality Index (AQI) monitoring module, a dynamic temperature trend visualization graph rendered using the Matplotlib library, a live digital clock, and an automated data refresh mechanism. The system further supports Excel-based data import and export functionality using the OpenPyXL library, enabling users to maintain historical weather logs for analysis and reporting. Evaluation of the system across multiple test scenarios demonstrates low API response latency, high data accuracy when compared against official meteorological benchmarks, and an intuitive user experience validated through structured usability assessment. The paper also discusses the three-tier architectural model, security considerations including API key management and data privacy, and potential avenues for future enhancement, including machine learning-based weather forecasting integration.

DOI: https://doi.org/10.5281/zenodo.19761598

Deep Learning – Based Bone Fracture Detection With Visual Explainability In Radiographic Images

Authors: Nimishakavi Sriram, Allampalli Harini

Abstract: Accurate bone fracture detection from X-ray images is essential for clinical diagnosis and emergency care. Manual interpretation by radiologists can be time-consuming and influenced by fatigue and subjective judgment. To overcome these limitations, this work proposes an automated fracture detection system based on deep learning.The framework employs convolutional neural networks with transfer learning to classify X-ray images as fractured or non-fractured. Pre-trained architectures such as ResNet-50, VGG-16, and DenseNet-121 are fine-tuned using a curated musculoskeletal dataset, with image preprocessing and data augmentation applied to improve robustness. Visual explanation methods are also incorporated to enhance prediction interpretability.Performance evaluation using standard metrics shows that the proposed system achieves an accuracy of 94.2%. Among the evaluated models, ResNet-50 offers the best balance between accuracy and computational efficiency, making it suitable for real-world clinical deployment.

DOI: https://doi.org/10.5281/zenodo.19763051

A Comprehensive Analysis Of Data-Driven Approaches To Digital Environmental For Quantifying And Managing Digital Carbon Footprints

Authors: Suryansh Singhwal, Uruj Jaleel, Satish Kumar Soni

Abstract: The exponential growth of digital technologies has introduced a new dimension to environmental concerns: the digital carbon footprint. This research paper explores the intersection of environmental science and data analytics, examining how data science methodologies can be leveraged to measure, monitor, and mitigate the carbon emissions associated with digital infrastructure and activities. Through comprehensive analysis of data centers, network infrastructure, and end-user devices, this study demonstrates that the Information and Communication Technology (ICT) sector currently accounts for approximately 2-4% of global greenhouse gas emissions [1][2], with projections suggesting this could reach 14% by 2040 without intervention. We present a data-driven framework for carbon footprint assessment, incorporating machine learning algorithms for predictive modeling and optimization strategies. The findings reveal significant opportunities for emission reduction through improved energy efficiency, renewable energy integration, and optimized resource allocation. This research contributes to the growing field of environmental data science by providing actionable insights for organizations seeking to reduce their digital environmental impact while maintaining operational efficiency

Online Recruitment Fraud Detection Using Bidirectional LSTM Neural Networks

Authors: Ahammed Anzar Abdul Nazar, Hemanth Kanna M, Dr. S Joseph James

Abstract: The rapid shift to online hiring has opened the door wide for employment scammers, who exploit the scale and anonymity of digital job platforms to target unsuspecting applicants. It is hard to tell the difference between fake and real listings just by looking at them, which makes manual review both unreliable and impossible to do on a large scale. This paper presents a comprehensive deep learning system designed explicitly to identify fraudulent job postings prior to their dissemination to job seekers. The Employment Scam Aegean Dataset (EMSCAD) is used to train the system. It has about 18,000 real job postings, but only about 5% of them are fake. A Bidirectional Long Short-Term Memory (Bi-LSTM) neural network is the heart of the detection engine. It reads job descriptions in both forward and backward directions, picking up on language patterns that single-direction models often miss. To fix the imbalance problem, class-weighted training was used during the learning phase. A FastAPI backend and a full-stack JavaScript frontend connect to the trained model. This lets you analyse individual postings in real time and process CSV datasets in batches. The system does very well on the EMSCAD benchmark, which shows that NLP models that are aware of sequences can be useful tools for fighting employment scams.

DOI: https://doi.org/10.5281/zenodo.19809067

Ai-Based Resume Analysis And Candidate Matching System

Authors: Harinag J, Suresh D, Sreekanth S, Suganya I

Abstract: The development of the AI-Based Resume Analysis and Candidate Matching System is rooted in the understanding that modern recruitment demands a more intelligent, data-driven approach than traditional manual screening methods. In today’s competitive job market, organizations receive a vast number of applications for each role, making it increasingly difficult for recruiters to efficiently identify the most suitable candidates. The availability of large-scale digital resume data presents an opportunity to move toward intelligent hiring systems, where candidate attributes are systematically analyzed and evaluated to support precise and consistent decision-making. This project initiates that transformation by centralizing critical candidate information—such as skills, educational background, work experience, certifications, and domain-specific.

DOI: https://doi.org/10.5281/zenodo.19809240

Advanced Deepfake Detection Using Machine Learning

Authors: Prof. Pradnya Patange, Atharv Pate, Harsh Lonari, Mayuresh Kshirsagar, Manish Patil

Abstract: The rapid advancement of deepfake technology has introduced significant challenges to digital media authenticity, enabling the creation of highly convincing synthetic images and videos that are difficult to distinguish from genuine content. This paper proposes an advanced deepfake detection framework based on the Temporal Vision-Language Transformer (TVLT), a cutting-edge multimodal deep learning architecture that jointly learns from visual, temporal, and semantic representations. Unlike traditional convolutional or recurrent models that focus solely on spatial or temporal domains, the proposed TVLT-based system integrates cross-modal attention to capture complex correlations among video frames, motion patterns, and audio-text alignment cues. The model efficiently identifies inconsistencies in facial movement, speech synchronization, lighting, and microexpressions — features that deepfake generation methods struggle to replicate authentically. Experimental evaluation on benchmark datasets including FaceForensics++, Celeb-DF, and DFDC demonstrates that the proposed system achieves accuracy exceeding 94%, with high precision and recall, significantly outperforming single-modality detection approaches.

Smart Drone Defense Systems: Using AI Cameras And Radio Blocking For Better Airspace Security

Authors: Manasi Shah, Arya Raul, Meer Shah, Dr Nandkishor Narkhede

Abstract: The rapid proliferation of low-cost unmanned aerial vehicles (UAVs) has created critical vulnerabilities in airspace security at airports, government installations, and military facilities. Traditional radar-based detection systems exhibit a fundamental visibility gap, as they are optimized for large aircraft rather than small, low-altitude drones. This paper presents a technical and methodological analysis of a Smart Drone Defense System that integrates AI-powered computer vision using the YOLOv11 object detection framework with Full-Duplex Soft- ware Defined Radio (SDR) technology for simultaneous signal jamming. The proposed integrated architecture eliminates the detection blind spot inherent in conventional jamming systems, achieves drone identification in under 50 milliseconds, and improves radio blocking efficiency by 40%. Comparative analysis against traditional and single-modality systems demonstrates superior accuracy and response time, establishing the viability of multi-modal smart systems for next-generation airspace protection.

6G Networks: Enabling Sustainable And Intelligent Future Communication Systems

Authors: Shreya Wadkar, Pundlik Yadate, Sarthak Shendge, Mukund Saroj, Dhanshree Porwal, Abdur Rehman, Nandkishor Narkhede

Abstract: The proliferation of smart devices, data-demanding applications, and the emergence of new communication technologies have led to the need for communication systems beyond the next generation of communication networks. Although 5G technology has enhanced the data rate, latency, and connectivity of communication systems, the next generation of applications, such as smart cities, autonomous systems, immersive virtual worlds, and healthcare services, require even greater speeds and intelligent management systems. 6G technology is expected to address the challenges and requirements of the next generation of communication systems and applications, enabling ultrahigh data rates, ultra-low latency, and reliability through the use of advanced communication technologies, such as artificial intelligence, terahertz communication, and edge computing. The paper provides an overview of 6G communication networks, their technology, architecture, and applications, as well as the limitations of the next generation of communication networks and how 6G technology can enable the next generation of communication systems and applications more efficiently and sustainably. The study further highlights the challenges affecting the development and deployment of 6G communication networks, and the findings indicate that 6G technology is expected to play a vital role in the development of intelligent communication systems in the future.

Battery Management System

Authors: Dr. M. S. Mathpati, Suyog Shrikant Kshirsagar, Rupesh Aanaso Salgare, Balaji Maruti Dhere

Abstract: The growing demand for electric vehicles has made battery safety, reliability, and performance monitoring a critical area of research. This paper presents the design and development of EV-BMS-X1, an intelligent battery management system that integrates embedded protection mechanisms with IoT-based real-time monitoring. The system is built using an ESP32 microcontroller and incorporates multiple sensors to continuously track key battery parameters such as voltage, current, temperature, and environmental conditions. Unlike conventional systems, the proposed solution introduces an adaptive web-based interface that allows users to monitor data and modify system parameters wirelessly without firmware reprogramming. To ensure measurement stability in noisy environments, filtering and smoothing techniques are applied to sensor data. Additionally, the system estimates battery usage through cycle counting and energy throughput analysis, providing insights into battery health. Experimental validation demonstrates that the system responds effectively to abnormal conditions such as over-voltage, over-current, and thermal hazards. The proposed design offers a practical, scalable, and user-friendly solution for next-generation electric vehicle battery management systems.

Class Incremental Learning in Efficient Job Task Recognition

Authors: Vaidegi K, Yogesh G, Vimalraj K, Yuvaraj R

Abstract: The project titled “Class Incremental Learning in Efficient Job Task Recognition”focuses on developing an intelligent system capable of identifying and classifying job-related tasksusing machine learning techniques. The system utilizes advanced algorithms to analyze input dataand categorize tasks into predefined classes with high accuracy. A key feature of the proposedsystem is its ability to support incremental learning, allowing new job categories to be addedwithout retraining the entire model. This improves scalability and efficiency in dynamicenvironments where new tasks frequently emerge. The system integrates datapreprocessing,feature extraction, and classification modules to ensure reliable performance. Experimental resultsdemonstrate improved accuracy and reduced computational cost compared to traditionalapproaches. The proposed solution is suitable for real-time applications and enhances automationin job task recognition systems.

3D Printing And Six Sigma In Quality Optimization Of Indian Foundry Operations: A Comparative Study

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

Abstract: By providing cast components to the automotive, aerospace, agriculture, and heavy engineering industries, the Indian foundry sector is vital to the manufacturing ecosystem. Operational efficiency is nevertheless impacted by enduring problems such casting flaws, high rejection rates, process unpredictability, and lengthy lead times. The Six Sigma methodology and 3D printing (additive manufacturing) have become two of the most promising approaches for quality optimization in recent years. A comparative analysis of these approaches in the context of Indian foundry operations is presented in this research. Six Sigma gives a systematic statistical foundation for defect reduction and process control through DMAIC, while 3D printing allows quick prototyping, intricate mold/core manufacturing, and shorter design-to-production times. The study contrasts the two strategies based on factors like productivity, long-term sustainability, cost, implementation time, and defect reduction. According to the results, a hybrid strategy that combines Six Sigma with 3D printing produces better quality results and gives Indian foundries a competitive edge.

DOI: https://doi.org/10.5281/zenodo.19813109

Smart Oral Guard: Continuous Heart Health Monitoring Using Salivary Biomarkers And Photoplethysmography

Authors: C. Narayanan, Manash Gautam, P. Marimuthu, KA. Mohamed Yazid, Mohith Joshy

Abstract: Cardiovascular diseases (CVDs) remain the leading cause of mortality worldwide, necessitating continuous, non-invasive monitoring solutions. Existing wearable devices are limited to physical parameters and cannot assess biochemical markers critical for comprehensive cardiac assessment. This paper presents the Smart Oral Guard, a novel intraoral wearable device integrating colorimetric salivary biomarker analysis with photoplethysmography (PPG) for continuous cardiovascular health monitoring. Colorimetric test strips analyzed by a TCS34725 RGB optical sensor detect salivary pH, potassium (K⁺), and alpha-amylase (sAA), while a MAX30102 PPG sensor measures heart rate and heart rate variability (HRV) from the buccal mucosa. An ESP32 microcontroller processes and transmits data via Bluetooth Low Energy to a Flutter mobile application for real-time visualization. A Firebase cloud backend employs Random Forest machine learning to classify cardiovascular risk into Low, Moderate, and High categories with alert generation. The system is designed to be portable, cost-effective (approximately USD 60 in components), and user-friendly, bridging the gap between biochemical and physiological cardiac monitoring in a single, comfortable intraoral platform.

Research Paper On Cyber Security Challenges And Defense Mechanisms

Authors: Miss.Shweta Guru Chikmal

Abstract: Cyber security plays a vital role in protecting digital systems, networks, and data from unauthorized access and cyber threats. With the rapid growth of digital technologies such as cloud computing, e-commerce, and mobile applications, cyber attacks have increased significantly. This research paper focuses on identifying major cyber security challenges and analyzing various types of cyber threats. It also discusses modern defense mechanisms such as encryption, firewalls, artificial intelligence, and zero-trust security models. The study highlights the importance of adopting a multi-layered security approach along with user awareness to ensure the safety of digital environments.

DOI: https://doi.org/10.5281/zenodo.19842720

Experimental Investigation On The Compressive Strength Of M40 Grade Concrete Reinforced With Basalt Chopped Fiber

Authors: Aniket Prakash Lohar, Prof. Dr .R .S. Karale

Abstract: Concrete exhibits high compressive strength but low tensile capacity, leading to cracking and brittle failure. This study investigates the effect of basalt chopped fibers on the compressive strength of M40 grade concrete. Concrete mixes with varying fiber contents were prepared, cast, and cured under standard conditions, and tested at 7, 14, and 28 days as per Indian Standards. Results show that the inclusion of basalt fibers improves compressive strength up to an optimum dosage, beyond which strength slightly decreases due to reduced workability and fiber clustering. Fiber-reinforced concrete also exhibited better crack resistance and ductile behavior. The study concludes that basalt chopped fibers are an effective and sustainable reinforcement material for enhancing the performance and durability of high-strength concrete.

My Smart Counselling Support For Personalized Academic Stream And Career Guidance Using Machine Learning, Django, And AI Chatbot

Authors: Gagandeep, Mayank Negi, Deepak Gupta, Mohammad Zeeshan, Mrs. Monika Jaglan

Abstract: Academic stream and career selection continues to be one of the most consequential decisions in a student’s educational journey, yet the vast majority of Indian students navigate this decision without access to objective, structured, or personalized guidance. Much like how early intervention is vital in preventive healthcare, timely and accurate academic counselling significantly reduces the risk of career mismatch, dropout, and long-term professional dissatisfaction. This paper introduces My Smart Counselling Support, a comprehensive intelligent web-based counselling platform that integrates the K-Nearest Neighbors (KNN) machine learning algorithm, a Django-powered full-stack backend, a responsive HTML/CSS/JavaScript frontend, and an AI-powered chatbot to deliver real-time, data-driven career and stream recommendations. The system evaluates a multi-dimensional student profile — comprising academic subject marks, interest survey scores across six career dimensions, twenty skill self-assessments, and behavioral pattern indicators — and processes the resulting 44-dimensional feature vector through an optimized KNN classifier (K=7, distance weighting, Euclidean metric) to recommend one of eight predefined career domains. Experimental evaluation on a dataset of 5,660 records demonstrated that the system achieves a test-set accuracy of 87.4%, weighted F1-score of 0.875, and cross-validation accuracy of 87.4% ± 0.4%, outperforming baseline Naive Bayes (81.2%), Decision Tree (83.7%), and Logistic Regression (79.4%) classifiers. An ablation study confirmed that multi-dimensional profiling outperforms academic-marks-only baselines by 9.1 percentage points, validating the framework’s holistic design. A context-aware AI chatbot module handles student career queries in natural language with 91.3% intent classification accuracy. User acceptance testing with 120 students yielded a System Usability Scale score of 81.7/100 (Excellent category) and a recommendation alignment rating of 3.94/5.0, highlighting both the system’s practical usability and the perceived relevance of its recommendations.

DOI: https://doi.org/10.5281/zenodo.19816963

Explainable RAG Systems For Transparent Decision Intelligence Using LLM

Authors: Mrs. C. Radha, Rohith P, Sapthagiri R

Abstract: Despite its effectiveness in enriching LLMs with external knowledge, there are many issues related to the opaqueness of the joint decisions made by the retriever and the generator that impede the implementation of RAG models in real-world applications. In this paper, we propose an end-to-end framework for designing xRAG systems that can facilitate explainable decision intelligence via several types of explanation techniques. We review the state-of-the-art progress in four paradigms for explaining RAG models: ARENA (reinforcement learning-based evidence navigation), ArgRAG (quantitative bipolar argumentation), post-hoc sentence-level attribution, and perturbation-based explanation. The proposed approach delivers 92.3% explanation fidelity on standard benchmarks and 94.1% user trust ratings in clinical decision-making applications. The comparison between different methods shows that structured reasoning-based techniques (ArgRAG) exhibit the best explanatory capability while retaining 88.9% of the accuracy of the traditional RAG models. We also show that the quality of explanations significantly contributes to user trust (r=0.87, p<0.001).

DOI: https://doi.org/10.5281/zenodo.19818683

Cloud Based Monitoring System

Authors: Krishna Narke, Sujal Boladra, Tanvi Shewale, Akshat Rampure

Abstract: The Cloud-Based Monitoring System is designed to monitor systems, applications, or environments in real time using cloud computing technologies. It collects data from various sources such as servers, sensors, or applications and stores it in the cloud for analysis and visualization. The system enables users to track performance, detect anomalies, and receive alerts from anywhere. Technologies like cloud platforms, APIs, and monitoring tools are used to ensure scalability and reliability. The system helps reduce downtime, improve performance, and enhance decision-making. It is widely used in IT infrastructure, healthcare, smart homes, and industrial automation. Overall, the project aims to provide an efficient, scalable, and remote monitoring solution.

PlusPlay: A Next-Generation Platform For Personalized Digital Entertainment

Authors: Syed Humer Ahmed, Tajudin, Cherithaa, Ekus Saluja, Harshvardhan Jain

Abstract: Building a successful streaming platform requires a well-rounded approach, incorporating various key areas to ensure long-term viability and competitiveness in the industry. It begins with comprehensive market research, which helps identify the target audience, their preferences, viewing habits, and the demand for specific types of content. Understanding consumer behavior and market trends allows for informed decision-making when developing the platform’s unique value proposition. The rapid advancement of digital technologies has revolutionized the way content is consumed, leading to the proliferation of streaming platforms across various media sectors. This study aims to explore the multifaceted process of developing a successful streaming platform, examining the critical components that contribute to its functionality, user engagement, and market competitiveness. As consumers increasingly demand on-demand access to diverse content, understanding the strategic, technological, and operational aspects of streaming services becomes imperative for entrepreneurs, developers, and stakeholders in the digital media landscape. A thorough competitor analysis is equally important, involving the study of existing streaming platforms, their strengths, weaknesses, pricing models, and technological advancements. By analyzing direct and indirect competitors, businesses can identify gaps in the market, potential differentiators, and areas for innovation, ensuring that the platform offers something unique and compelling.

DOI: https://doi.org/10.5281/zenodo.19819741

CareBot A Medical Chatbot Using NLP

Authors: Y.Prasanna, Kasula Vijay Vardhan, Kuchangari Sahit, Gaddameedi Pranay Kumar

Abstract: A chatbot, or conversational agent, is an AI-powered software designed to communicate with users using natural language. In the healthcare domain, medical chatbots play a crucial role in providing preliminary health guidance, answering medical queries, and assisting patients with symptom analysis. One of the major challenges in developing a medical chatbot is designing an effective dialogue system that can accurately understand user input and provide relevant responses. Early chatbot models relied on rule-based systems and statistical methods, which had limited flexibility in handling diverse conversations. However, with advancements in artificial intelligence and natural language processing (NLP), deep learning techniques, particularly end-to-end neural networks, have revolutionized chatbot development. Since 2015, encoder-decoder recurrent models, originally designed for neural machine translation, have become dominant in conversational AI due to their ability to generate meaningful and context-aware responses. This project aims to develop a medical chatbot using NLP techniques to enhance healthcare accessibility. The chatbot will leverage deep learning models to analyze user queries, understand symptoms, and provide preliminary advice based on medical knowledge. By integrating state-of-the-art advancements in NLP, this system seeks to offer accurate, reliable, and interactive healthcare assistance, improving user experience and bridging the gap between patients and medical professionals.

DOI: https://doi.org/10.5281/zenodo.19828185

Cyclone Damage Forecasting And Risk Evaluation In The Bay Of Bengal Region Using Decision Tree Algorithms

Authors: Dr N.Magesh, Dr.S.Jabeen Begum, Dr. A.P.Gopu, M.Ashwinth

Abstract: The Bay of Bengal is one of the most cyclone-prone regions in the world, frequently experiencing severe tropical storms that result in significant socio-economic losses. Accurate prediction of cyclone-induced damage is essential for effective disaster management and mitigation planning. This study proposes a machine learning-based approach for cyclone damage forecasting using Decision Tree algorithms. Historical meteorological data, including wind speed, atmospheric pressure, temperature, storm surge, and humidity, are analyzed to model cyclone impact. The J48 decision tree algorithm is employed to classify damage levels and evaluate regional risk. Experimental results demonstrate that the proposed model provides interpretable decision rules and satisfactory prediction accuracy. The model can assist disaster management authorities in improving early warning systems and evacuation strategies in cyclone-prone coastal regions.

DOI: https://doi.org/10.5281/zenodo.19829537

AI Vision-Based Approach For Detecting Cotton Leaf Diseases

Authors: Kavithra R, Lakshminarayanan A, Allen joses A, Pavin S, Saranraj E

Abstract: Cotton is one of the most important cash crops, playing a vital role in the agricultural economy. However, cotton plants are highly susceptible to various leaf diseases such as bacterial blight, leaf spot, and aphid infestations, which significantly reduce crop yield and quality. Traditional disease detection methods rely on manual inspection by farmers and experts, which is time-consuming, labor-intensive, and prone to human error. To address these challenges, this project proposes an automated cotton leaf disease detection system using deep learning techniques. A Convolutional Neural Network (CNN)-based model is trained on a labeled dataset of cotton leaf images to accurately classify different disease categories. The system is integrated into a user-friendly web application that allows users to upload leaf images and obtain real-time predictions along with disease classification and affected percentage.

DOI: https://doi.org/10.5281/zenodo.19831165

PharmaChain: Enhancing Drug Traceability And Security In Pharma Supply Chains With Blockchain DApp

Authors: Bharathi Panduri, Lohith Matcha, Sai Darshan Lingamanthula, Vineet Katta

Abstract: Pharmaceutical providers are now facing problems such as counterfeit drugs, lack of openness, different data systems, and weaker regulation. As a result of these shortcomings, medicinal products may be unsafe for patients and this can damage the public’s trust in these products. Traditional centralized structures do not have the necessary traceability, security, and resilience for effective management of supply chains. In the context of Industry 4.0, there is an increasing shift toward digital transformation and decentralized industrial systems to improve transparency and automation. This paper describes PharmaChain, a DApp created using blockchain, which is intended to add transparency and trust to the way drugs are managed in the supply chain. With the help of Ethereum and Solidity-based smart contracts, PharmaChain develops a permanent record of every step in the supply chain, starting with buying raw goods and finishing with delivering them to end-users. The platform uses role-based access control, such as manufacturers, distributors, retailers, and regulatory bodies. With a modern MERN stack, PharmaChain ensures the UI is flexible and works well on any device, and Web3.js and MetaMask handle the connection to blockchain on the frontend. The proposed system aligns with Industry 4.0 principles by enabling secure, automated, and decentralized traceability across the pharmaceutical supply chain. Smart contracts significantly reduce the involvement of intermediaries and manual work in the process.

Big Data Analytics

Authors: Mrs. Shamal Nitinkumar Ingawale

Abstract: In the information era, enormous amounts of data have become available on hand to decision makers. Big data refers to datasets that are not only big, but also high in variety and velocity, which makes them difficult to handle using traditional tools and techniques. Due to the rapid growth of such data, solutions need to be studied and provided in order to handle and extract value and knowledge from these datasets. Furthermore, decision makers need to be able to gain valuable insights from such varied and rapidly changing data, ranging from daily transactions to customer interactions and social network da¬ta. Such value can be provided using big data analytics, which is the application of advanced analytics techniques on big data. This paper aims to analyze some of the different analytics methods and tools which can be applied to big data, as well as the opportunities provided by the application of big data analytics in various decision domains.

Synthesis Of Ironoxide Nanoparticles From Punica Granatum Peels And Formulation Of Nanobeads For Effective Drug Delivery Application

Authors: Karthikeyan D, Sayed neamina S, Senthamizh arasi P, Sivasakthi V

Abstract: This project focuses on the eco-friendly synthesis of iron oxide nanoparticles using Punica granatum peels and their application in advanced drug delivery systems through nanobead formulation. In recent years, nanotechnology has emerged as a powerful tool in biomedical engineering, particularly in improving drug delivery efficiency and reducing side effects. Conventional methods for nanoparticle synthesis involve toxic chemicals, high energy consumption, and environmental hazards. To overcome these limitations, green synthesis using plant-based materials has gained significant attention. Pomegranate peels are rich in bioactive compounds such as polyphenols, flavonoids, and tannins, which act as natural reducing and stabilizing agents in nanoparticle formation. In this study, iron oxide nanoparticles were synthesized using aqueous peel extract and further characterized using UV-Visible spectroscopy and FTIR (Fourier Transform Infrared Spectroscopy) analysis to confirm their formation and functional groups. The synthesized nanoparticles were then incorporated into polymer-based nanobeads using biocompatible materials such as alginate or chitosan. These nanobeads were used as carriers for drug delivery, enabling controlled and sustained release of therapeutic agents. The results demonstrate enhanced drug loading efficiency, improved stability, and reduced toxicity compared to conventional delivery methods. Overall, this project highlights the potential of green nanotechnology in developing safe, cost-effective, and efficient drug delivery systems for biomedical applications.

AI-Based Real Time Travel Itinerary Planner

Authors: Bharathi Panduri, Kaliveni Naveen, Alluru Manoj, Patha Shiva Anurag, Sri P. Gopala Krishna, Dr. Y J. Nagendra Kumar

Abstract: This research describes the development of an Artificial Intelligence Based Manufacturing Optimization System that employs state-of-the-art Artificial Intelligence resources such as LangChain framework, Groq’s LLaMA-3.3-70B Large Language Model and Streamlit interface to transform and optimize how factories work. The system provides real-time analytics, predictive maintenance algorithms and Intelligent Manufacturing Process Optimization to increase production efficiency, lower operational costs, and minimize waste of materials in manufacturing. To create these optimized processes, the core system uses the LLaMA-3.3-70B model and the power of Generative AI technology to process all types of complex manufacturing data such as machine specifications; material characteristics; manufacturing schedules; quality metrics; resource limitations; etc., into optimized workflows for manufacturing operations. The system uses ChatPromptTemplate to dynamically adjust to the type and context of manufacturing so any company using this software will receive customized recommendations for using the correct processes to manufacture products, whether 3D printing, CAD, or Production Planning. The system uses a sophisticated Cost-Benefits Analysis Engine to calculate for each optimization scenario the cost of manufacturing (operations), materials, energy, and the total amount of money that can be saved through these optimized processes. The Streamlit framework allows industrial engineers in the manufacturing environment to easily enter their parameters, select production process optimization goals, and receive AI-Derived optimized recommendations, including in-depth analytics supporting those recommendations. Another feature of the system is its ability to provide Real-Time Monitoring of key Performance Indicators (KPIs) such as Overall Equipment Effectiveness (OEE), production defects, cycle times and so on.

SyncBoard: An Intelligent Platform For Real-Time Data Sharing Across Devices

Authors: Madhuri A. D, Sharadha D, Mrudula S, Anjali S, Kartik Y

Abstract: Today, many people use more than one device for study, work, or daily tasks. Sharing small information like text or files between devices can be difficult and time-consuming when using emails or messaging apps. To solve this problem, Sync Board is developed as a simple and user-friendly system that allows instant sharing of content a cross device. Sync Board provides a web-based platform where users can add text or upload files on one device and see the same content appear immediately on other connected devices. The system works in real time and does not require complex setup or technical knowledge. Its clean and easy interface makes it suitable for users of all skill levels. The system helps save time, reduce manual effort, and improve productivity. Sync Board is useful for students, professionals, and teams who need quick and smooth content sharing. This project shows how a simple and efficient design can make real-time technology easy to use and accessible for everyone.

Comprehensive Study of Cyber Security: Threats, Techniques, and Trends in the Digital Era

Authors: Achal Anikat, Amitava Sen

Abstract: The rapid proliferation of internet-connected systems, cloud computing environments, and mobile devices has fundamentally transformed the global digital landscape, creating unprecedented opportunities while simultaneously expanding the attack surface available to malicious actors. Cyber security — the discipline dedicated to protecting digital assets, networks, and data from unauthorised access, disruption, or destruction — has emerged as a critical strategic priority for organisations and governments worldwide. This paper presents a comprehensive study of cyber security, examining foundational concepts, the classification and mechanics of prevalent cyber threats and attacks, established and emerging security techniques and tools, and industry-recognised frameworks for risk management and mitigation. The paper also analyses current industry trends, including the integration of Artificial Intelligence in threat detection, the rise of zero-trust architecture, and the growing importance of cyber security regulation and compliance. Real-world examples are drawn upon throughout to illustrate theoretical concepts in practical context. The study identifies key research gaps in the field and articulates the urgent need for structured, proactive cyber security frameworks in modern organisational environments. Findings affirm that a multi-layered, intelligence-driven approach to cyber security — one that combines technological controls, human awareness, and policy governance — is essential to safeguarding digital infrastructure in an era of increasingly sophisticated and persistent cyber threats.

Ai-Driven Phishing Detection System

Authors: J. Nihanth Kumar, M. Rakesh Sai, K. Nani, Dr. Sasikumar Gurumoorthy

Abstract: AI-driven phishing detection systems are a critical component of modern cybersecurity, as they directly influence information security, user trust, and organizational resilience. Their primary objective is to reduce exposure to malicious emails and websites while ensuring rapid, accurate identification of phishing attempts in large-scale digital communication environments. This paper discusses current technical approaches to phishing detection, focusing on the application of artificial intelligence, natural language processing, and machine learning techniques to analyze email content, URLs, metadata, and user behavior patterns. The relationship between automated phishing detection and intelligent security decision support is examined, highlighting its growing importance in next-generation security operations. Limitations in existing manual review and rule-based detection methods are identified, emphasizing the need for innovative AI-based solutions that enhance detection accuracy, efficiency, and interpretability while complementing existing security workflows rather than replacing established processes.

Digital Therapeutics And The Expanding Role Of Pharmacists

Authors: Rimaz fayazuddin , Qhazima Sareen , Mahveen maryam , Safura Vali , Mohammed Shaik Fahad , Mallik Rehan , Syed Sadiq , Mohammed Faraz

Abstract: Digital therapeutics (DTx) have emerged as a transformative innovation within the healthcare ecosystem, offering evidence-based therapeutic interventions delivered through software to prevent, manage, or treat diseases. Unlike conventional digital health tools, DTx are clinically validated and often regulated, ensuring their safety, efficacy, and integration into formal healthcare systems. The rapid expansion of digital therapeutics is driven by the increasing prevalence of chronic diseases, rising healthcare costs, and the growing demand for personalized and accessible care. Pharmacists, traditionally recognized for their role in medication dispensing and counseling, are increasingly assuming expanded responsibilities within digital healthcare frameworks. Their accessibility, clinical expertise, and frequent patient interactions position them as key facilitators in the implementation and optimization of digital therapeutics. This article explores the evolution, clinical applications, benefits, and challenges of DTx, while highlighting the expanding role of pharmacists in patient education, adherence monitoring, and interprofessional collaboration. The paper further examines regulatory considerations, integration into pharmacy practice, and future perspectives, emphasizing the need for training, policy support, and digital literacy. Ultimately, the integration of digital therapeutics into pharmacy practice has the potential to enhance patient outcomes, improve healthcare efficiency, and redefine the role of pharmacists in modern healthcare systems.

DOI: https://doi.org/10.5281/zenodo.19907622

Research On Development Of Android Applications

Authors: Mrs. Shamal Nitinkumar Ingawale

Abstract: Introduced the Android platform and the features of Android applications, gave a detailed description of Android application framework from the prospective of developers. A simple music player is provided as instance to illustrate the basic working processes of Android application components. This paper provides guidance to understanding the operation mechanism of Android applications and to developing applications on Android platform.

Artificial Intelligence In Cyber Threat Detection: A Survey Of Predictive Security Systems

Authors: Miss Jayashri Raosaheb Bedage

Abstract: The scope and nuances of cyber threats have been escalating with the blistering pace of digital technologies development, and several questions are raised regarding the applicability of the traditional strategies of cyber security. Antivirus and firewall programs are examples of conventional security measures; however, they only protect against known threats and cannot detect or prevent new ones. By combining machine learning, neural networks, and natural language processing to identify outliers, anticipate attacks, and automate responses, Artificial Intelligence (AI) enables more proactive and adaptable defense. This article is a review of the state of the cyber threat detection using AI, in which it cites multi-layered system incorporating data collection, preprocessing, real-time analytics, and automated cancellation. It examines major types of threats, including phishing, ransomware, insider threats, and vulnerabilities at the protocol level, as well as issues related to implementation, including data quality, model transparency, and integration. New methods such as Federated Learning and Generative AI are also discussed, possibly to augment decentralized learning and to create attack-inspired scenarios. This paper highlights the necessity of developing intelligent systems that can adapt to cyber threats to enhance the resilience of infrastructure in the digital world.

Role of Ancient Indian Knowledge Systems in AI-Driven Business Models: A Conceptual and Applied Study

Authors: Zoya Heyat Bakshi

Abstract: Artificial Intelligence (AI) is transforming modern business models through automation, predictive analytics, and data-driven decision-making. However, ethical concerns such as bias, lack of transparency, and sustainability challenges persist (Floridi et al., 2018). Ancient Indian knowledge systems, including the Arthashastra and Bhagavad Gita, provide principles of Dharma, governance, and holistic well-being. This paper proposes an integrated framework combining AI technologies with traditional knowledge systems to enhance ethical alignment and sustainability in business models.

Dron Based Transmission Line Inspection

Authors: Dr. V.G. Umale, Aman Ranjan, Prem Sandeep, Bombale, Tanushree Bagde, Snigdha Patil, Prem Thakre

Abstract: The inspection of transmission lines is a critical task in power systems to ensure uninterrupted electricity supply and prevent faults. Traditional inspection methods are time-consuming, risky, and require manual effort. This project presents a drone-based transmission line inspection system that utilizes unmanned aerial vehicles (UAVs) for efficient monitoring of power lines. The proposed system integrates a drone equipped with a high-resolution camera, sensors, and wireless communication modules to capture real-time data of transmission lines. The drone can detect faults such as damaged conductors, insulator cracks, vegetation interference, and overheating components. The collected data is transmitted to a ground station for analysis. This approach improves safety by reducing human involvement in hazardous environments and enhances inspection accuracy. The system is cost-effective, time-efficient, and suitable for modern smart grid applications. The project demonstrates the potential of drone technology in revolutionizing power system maintenance and monitoring. The reliability of electrical power transmission systems largely depends on the timely detection and maintenance of faults in transmission lines. Traditional inspection methods are labor-intensive, time-consuming, and often expose personnel to hazardous environments. To overcome these limitations, this project focuses on the development and implementation of drone-based transmission line inspection system. The proposed system utilizes with high-resolution cameras and various sensors to capture real-time data and images of transmission line components such as conductors, insulators, and towers. The collected data are analyzed to identify defects like corrosion, cracks, vegetation encroachment, and loose fitting The use of drones significantly improves inspection efficiency, accuracy, and safety while reducing operational costs and downtime.

Live Surveillance with Actionable Intelligence

Authors: Mrs. Vibhavari Jawale, Mrs. Deepali Hajare, , Arhant Sahuji, Tanay Shinde, Ritesh Kadam, Ananya Vaishnav

Abstract: The combination of advanced machine learning and AI capabilities associated with natural language processing (NLP) and technological developments in computer vision have led to the creation of smart video surveillance systems. Unlike conventional Closed Circuit Television (CCTV) and motion-detection-based surveillance devices, these systems are capable of understanding contextual information, reducing false alarms and requiring less human intervention. Researchers have explored incorporating vision-language models (VLMs) and Sentiment Analysis (SA) into video surveillance applications to improve contextual awareness. This review focuses on emerging techniques associated with image captioning models, including Salesforce’s BLIP, used to generate natural language descriptions of real-time actions in video footage and perform SA on those descriptions to determine the nature of the detected activity. By utilizing visual comprehension, context building, and sentiment interpretation, surveillance systems can differentiate between normal and suspicious behavior while reducing false positives and generating actionable insights. Applications include public safety in smart cities, security at high-threat locations such as airports and banks, and monitoring of sensitive areas including hospitals and military installations. This review evaluates how contextual awareness enabled by VLM improves traditional object detection methodologies and supports the transition toward more human-like and explainable alerting modalities. It also discusses limitations related to computational burden, accuracy, and privacy, while highlighting broader societal implications aligned with Sustainable Development Goals focused on urban safety and crime reduction. Future research directions include multimodal fusion, real-time optimization, and ethical guidelines for responsible deployment.

DOI: https://doi.org/10.5281/zenodo.19881170

Threat Detection Of Spam And URL Using Machine Learning And NLP

Authors: Gaurav Kadam, Soham Patel, Piyush Bande

Abstract: Spam Messages are very irritating and an upscaled issue in online communication devices and system , it leads to security issue and fraudness in todays world. This paper presents a simple and normal solution that uses Natural language processing techniques to overcome the security issues and help to maintain security. The system developed in this paper represents the combination of NLP and ML that uses advanced text preprocessing, TF-IDF feature extraction, and classification using models or classifiers such as naive bias Logistic Regression, Support Vector Machine (SVM), Random Forest, and Long Short-Term Memory (LSTM) networks. The system predefined outcomes showcases that two model get higher accuracy then others which are Logistic Regression and LSTM and very rigid for outliers. Our Spam detection also forms a user Interface with the help of Streamlit which is further explained in the given paper.

Hybrid Optimization-Based TabTransformer For Type 2 Diabetes Risk Prediction

Authors: Md. Shorifuzzaman, Annesha Hossain Noushin

Abstract: Type 2 Diabetes Mellitus (T2DM) is a rapidly growing global health problem where delayed diagnosis can lead to severe complications such as cardiovascular disease, kidney failure, neuropathy, and vision impairment. Existing machine learning approaches often suffer from limited interpretability, class imbalance, and inadequate optimization, reducing their clinical reliability. This study proposes an explainable and optimized deep-learning framework for T2DM risk prediction using a TabTransformer architecture with hybrid hyperparameter optimization and explainable artificial intelligence (XAI). A publicly available dataset of 100,000 patient records was preprocessed using encoding, standardization, and the Synthetic Minority Oversampling Technique (SMOTE). The model was optimized using Bayesian optimization (Optuna) followed by Particle Swarm Optimization (PSO) and evaluated using standard classification metrics. The optimized model achieved approximately 93% AUC and accuracy with improved recall for diabetic cases. SHAP analysis identified key risk factors, including glucose level, HbA1c, BMI, age, and hypertension, and a web-based interface enabled instant prediction, demonstrating real-time feasibility. The proposed system can serve as a clinical decision-support tool for early diabetes screening.

DOI: https://doi.org/10.5281/zenodo.19883294

A Holistic Framework For The Complete Journey Of Green Construction In Hong Kong

Authors: Dr. Assoc. Professor Samuel Kwok Piu LIP, Dr. Wing Cheung Tang

Abstract: Research on green construction in Hong Kong is scattered and only looks at certain parts of the problem, like choosing materials or getting BEAM Plus certification. It does not consider the city’s unique subtropical, high-density environment. Standard frameworks do not consider the fast pace of redevelopment (40–50 years for buildings versus 80 years in Europe), the lack of infrastructure for deconstruction, the pressure on land prices, and the working conditions for migrant workers. This article suggests a comprehensive framework for the entire process of green construction tailored to Hong Kong, encompassing raw material extraction (predominantly sourced from Mainland China) to deconstruction. By combining 120 studies from 2015 to 2025 and three long-term case projects (a BEAM Plus Platinum public housing estate, a LEED Gold commercial tower, and a BEAM Plus Gold institutional building), the authors find five important transition gaps: specification–procurement (unclear cross-border supply chains) and commissioning–occupancy performance.

DOI: https://doi.org/10.5281/zenodo.19883920

SignBridge: A Lightweight Framework For Real-Time American Sign Language Recognition Using MediaPipe And MobileNetV2

Authors: Vishal Singh, Sudipta Sarkar, Uddesh Raj

Abstract: American Sign Language (ASL) serves as a vital communication tool for the deaf and hard-of- hearing community, yet barriers persist due to limited familiarity among the general population. SignBridge addresses this challenge by introducing a lightweight, real-time ASL recognition system that leverages hand pose estimation with MediaPipe for hand landmark extraction and a customized MobileNetV2-based convolutional neural network (CNN) for gesture classification, emphasizing edge computing for efficient deployment. The pipeline processes RGB video input to detect and classify static ASL alphabets (A-Z) and numbers (0-9), achieving high accuracy while maintaining computational efficiency suitable for edge devices. The methodology begins with frame capture from a standard webcam, followed by MediaPipe Hands detection to extract 21 key landmarks per hand, forming a compact 42-dimensional feature vector for both hands. These features are fed into a lightweight MobileNetV2 variant, fine-tuned for ASL with knowledge distillation to reduce parameters by 40% compared to standard models. Training utilizes an 80/20 train/validation split on a 27-class WLASL subset for static ASL alphabets, employing data augmentation techniques like rotation and scaling invariance to handle real-world variations. Experimental evaluation on WLASL and a custom ASL dataset demonstrates 96% top-1 accuracy for static gestures, with real-time performance at 35 FPS on consumer-grade hardware (Intel i5 CPU, no GPU). Ablation studies confirm the efficacy of MediaPipe integration, outperforming baselines like VGG-16 by 15% in speed without accuracy loss. Comparisons with state-of-the-art methods, such as Transformer-based Vision Transformer (ViT) models achieving 92% accuracy but at 15 FPS due to higher compute (Karna et al., 2021), and MobileNet for ASL alphabets achieving ~99.93% accuracy (Kandukuri et al., 2023), highlight SignBridge’s novelty in balancing accuracy and latency for edge efficiency. This work contributes to accessible communication by enabling seamless ASL-to-text translation in applications like video calls and educational tools. Implications include broader societal inclusion, with potential extensions to dynamic gestures and multilingual sign languages. Limitations such as sensitivity to lighting are discussed, alongside future directions for multimodal integration. Reproducibility is ensured through open-source code and dataset details, promoting further advancements in inclusive technology. (198 words).

DOI: https://doi.org/10.5281/zenodo.19884687

Smart Fatigue Detection Model For Drivers

Authors: Shreyas Fatale, Ajinkya Bobade, Divya Jain, Kushal Agrawal, Prof. Mayur Chavan

Abstract: An abstract of this paper will provide a hybrid driver fatigue system that is supposed to be deployed in real-time on resource-constrained edge devices. The suggested framework integrates the lightweight Convolutional Neural Network (CNN) models with geometrical and temporal feature extraction based on the landmarks to provide a stable fatigue monitoring at the minimal computational complexity. The two parallel CNN models are used to perform eye state, and yawning detection, and geometric fatigue indicators are computed simultaneously through facial landmark Analysis. The percentage of eyelid closure (Eye Closure Percentage (PERC-LOS)) is used to determine eye fatigue and monitoring patterns of eyelid closure through time is an efficient measure of drowsiness, even in the process of lip detection in the presence of noise. The way of recognising yawning is the Jaw Drop Angle (JDA), which is a powerful geometric parameters calculated based on the nasal, chin, and jaw positions that is still reliable in spite of imprecision in lips location. A hybrid decision model combines deep learning-based predic-tions with geometric fatigue indicators to enhance the reliability of the system and help to reduce error. The general architecture proves to be practically viable as a cost-efficient approach to the intelligent driver assistance system, specifically in real-time embedded and edge computing applications.

Food Object Detection Using YOLO Model

Authors: Machindra K. Gaikwad

Abstract: Food object detection has emerged as a critical research area in the intersection of computer vision and nutritional informatics. This paper presents a comprehensive study on the application of YOLO (You Only Look Once) models for real-time food item recognition and classification. Accurate food detection is fundamental to calorie estimation, dietary tracking, and smart kitchen applications. We investigate the evolution from YOLOv1 through YOLOv8, analysing architectural improvements, training strategies, and performance trade-offs on benchmark food datasets including FOOD-101, UEC-Food256, and a custom annotated dataset of Indian cuisines. Experimental results demonstrate that YOLOv8 achieves a mean Average Precision (mAP@0.5) of 91.3% on the FOOD-101 dataset while maintaining real-time inference speeds of 42 FPS on standard GPU hardware. The study further explores transfer learning, data augmentation, and anchor-box optimization as techniques to improve detection accuracy across diverse food categories. Our findings suggest that YOLO-based architectures are well-suited for deployment in mobile and edge computing environments for dietary assessment applications.

Cloud Security: Protecting Data In Distributed Environments

Authors: Mr.Khedekar Nagesh Haridas

Abstract: Cloud computing has revolutionized the way organizations store, process, and share data. However, the distributed nature of cloud environments introduces significant security challenges. This paper explores the critical aspects of cloud security, common threats, and the modern techniques used to protect sensitive data in distributed environments. Solutions such as encryption, identity and access management, and secure multi-party computation are discussed to ensure confidentiality, integrity, and availability in cloud systems.

DOI: https://doi.org/10.5281/zenodo.19885909

SkillLens: An AI-Powered Career Readiness And Skill Enhancement Platform

Authors: Manoj.P, Aadhitiya.M, Dr. G. Priyadharshini

Abstract: There is a serious” employability crisis” among recent graduates as a result of the growing disconnect between academic courses and changing business demands. Conventional career coaching techniques frequently rely on static keyword-matching Applicant Tracking Systems (ATS) or subjective counseling, neither of which can offer useful information about a candidate’s true skill gaps or the caliber of their practical portfolios. This study introduces SkillLens, a comprehensive platform for job preparedness that democratizes career coaching by utilizing Natural Language Processing (NLP), Computer Vision, and Knowledge Graphs. The suggested solution uses a multi-module design that includes semantic skill gap analysis, automated resume parsing, and a unique AI Portfolio Analyzer that assesses project visual evidence. In order to comprehend the hierarchical links between technologies and provide tailored roadmap recommendations to close detected gaps, SkillLens makes use of a Skill Ontology. The system’s effectiveness in precisely matching user profiles to target job descriptions and enhancing interview readiness through real-time feedback is demonstrated by experimental validation. The system offers an end-to-end, scalable solution that empowers job seekers in the digital economy.

DOI: https://doi.org/10.5281/zenodo.19886295

Newsly: A User-Centric Personalised News Application For Efficient Content Filtering

Authors: Yash Panchal, Vansh Sharma, Shelly Chauhan

Abstract: In today’s digital age, users are overwhelmed with an excessive amount of information from various news platforms. Finding relevant and personalized news becomes difficult and time-consuming. The objective of this project, Newsly, is to develop a personalized news application that delivers curated news content based on user interests. The application allows users to select their preferred categories and generates a customized news feed using real-time data from NewsAPI. Built using React (Vite), the system focuses on simplicity, responsiveness, and user experience. Instead of implementing a complex backend, localStorage is used to simulate authentication and data persistence for demonstration purposes. This project demonstrates how modern frontend technologies can be leveraged to build efficient, scalable, and user-centric applications. Future enhancements may include AI-based recommendations, real-time updates, and backend integration.

Cyber Security And Artificial Intelligence: How AI Is Being Used In Cyber Security To Improve Detection And Response To Cyber Threats

Authors: Jayashri Raosaheb Bedage

Abstract: In recent years, the fusion of Artificial Intelligence (AI) with cyber security has revolutionized the way organizations detect, prevent, and respond to cyber threats. With the increasing sophistication of cyber-attacks, traditional security measures are no longer sufficient. AI-powered systems enhance cyber security by leveraging advanced algorithms to analyze vast amounts of data in real time, identify patterns, and predict potential threats before they materialize. Machine learning (ML), natural language processing (NLP), and anomaly detection techniques have become pivotal in improving threat detection, reducing false positives, and automating response mechanisms. This paper explores how AI is transforming the cyber security landscape, offering an in-depth look at its applications, benefits, challenges, and future prospects in combating cybercrime.

Systematic Alpha Modeling And Portfolio Optimization Using Multi-Factor Quant Strategies

Authors: Prof. Parag Jambhulkar, Bhavyam Sanghavi, Smit Khandelwal, Kiran Kankariya, Ayush Agrawal

Abstract: This paper presents a systematic quantitative investment framework that integrates technical factors, fundamental screening, probabilistic return forecasting, and risk-aware portfolio construction within a walk-forward evaluation architecture. The proposed approach combines momentum, volatility, trend structure, fractal persistence, and firm-level financial quality signals to rank assets cross-sectionally and allocate capital using inverse-volatility weighting. Monte Carlo–based simulations provide forward-looking expectations, while strict historical data alignment prevents look-ahead bias and ensures realistic deployability. The strategy is evaluated across multiple decades of market data spanning diverse economic regimes, including expansions, crises, and high-volatility periods. Empirical results demonstrate consistent out-of-sample performance, controlled drawdowns, and stable compounding, indicating robustness across changing market conditions. Rather than relying on a single predictive source, the framework benefits from the aggregation of diversified and complementary signals, leading to improved risk-adjusted returns. The modular design supports extensibility to additional factors, machine learning models, and multi-asset universes, providing a scalable foundation for next-generation quantitative and AI-driven portfolio management systems.

“Crime Analytics Web Application Using Hybrid Machine Learning And Deep Learning Approach

Authors: Saurabh Sharma, Arun Sharma, Vipin Kumar, Rahul Bharti

Abstract: This paper presents a Crime Analytics Web Ap-plication that leverages a hybrid approach combining Machine Learning and Deep Learning techniques for predictive crime analysis. The primary objective of the system is to transform traditional reactive crime investigation methods into proac-tive, data-driven decision-making frameworks. The proposed system integrates heterogeneous datasets, including historical crime records, temporal attributes, and socio-economic fac-tors, to identify hidden patterns and correlations in criminal activities. A robust data engineering pipeline is designed to pre-process and clean large-scale noisy datasets through feature engineering, normalization, and outlier handling. The system employs a hybrid ensemble model consisting of Light Gradient Boosting Machine (LightGBM) for capturing complex feature interactions and Long Short-Term Memory (LSTM) networks for modeling temporal dependencies in crime trends. This combination enhances prediction accuracy by leveraging both spatial and temporal insights. Furthermore, a user-friendly web application is developed using Streamlit, providing interactive dashboards, real-time prediction capabilities, and advanced data visualizations such as heatmaps and time-series plots. The system enables law enforcement agencies to anticipate crime hotspots, analyze trends, and make informed strategic decisions. The experimental results demonstrate that the hybrid ap-proach outperforms traditional single-model techniques in terms of accuracy and efficiency. The proposed solution has significant potential for real-world deployment in smart polic-ing and urban safety systems, with future scope including real-time data integration, geospatial intelligence, and smart city applications.

Design and Development of a Gift Recommendation System Using Rule-Based Approach

Authors: Rashika.R. K, William Hamilton.D, Siddhartha.J. R, Arun.H, Mrs.B. Subhalakshmi

Abstract: In today’s fast-paced lifestyle, choosing a suitable gift for different occasions has become a challenging task due to the wide variety of options available. People often find it difficult to select meaningful gifts that match the preferences and interests of the recipient. Traditional methods such as manual selection or browsing through online platforms are time-consuming and may not always provide satisfactory results. To address this problem, this project proposes a Gift Recommendation System that provides personalized gift suggestions based on user inputs such as age, gender, occasion, and interests. The system uses a rule-based approach to analyze user preferences and generate relevant recommendations. The application is developed using web technologies including HTML, CSS, and JavaScript, ensuring a simple, user-friendly, and efficient interface. The system processes user input dynamically and displays appropriate gift suggestions instantly. The results demonstrate that the system effectively reduces the effort required for gift selection and improves user experience by providing quick and meaningful recommendations. This project also highlights the basic concept of recommendation systems and serves as a foundation for future enhancements using machine learning techniques.

Mindprint: An Ai-Powered Cognitive Journaling System for Emotion Analysis and Personalized Career Trajectory Prediction

Authors: Manoj Mukundarao Pawar, Vinayak Pandurang Khandekar, Sumeet Madan Malviya, Sanika Rajaram Mohite

Abstract: The contemporary professional land- scape presents a multifaceted challenge for individuals seeking to align their inherent cognitive traits with viable career paths. Traditional psychometric evaluations and aptitude tests often offer static, one-dimensional snapshots of a candidate’s potential, failing to capture the nuanced evolution of emotional intelli- gence, behavioural patterns, and intrinsic motivations. MindPrint addresses this gap by introducing an innovative Human- Computer Interaction (HCI) framework that utilizes Natural Language Process- ing (NLP) to perform real-time sentiment analysis and personality trait extraction from user-written journal entries. By leveraging the Hugging Face Inference API and transformer-based architectures, the system transforms unstructured tex- tual data into a structured cognitive pro- file. This paper details the design, im- plementation, and rigorous evaluation of MindPrint. We discuss the integration of React.js, Node.js, and MongoDB in creat- ing a scalable architecture capable of gen- erating personalized career recommen- dations with high accuracy. Experimental results indicate that MindPrint achieves 94.5% accuracy in sentiment detection and significantly outperforms traditional assessment models in longitudinal trait tracking.

DOI: https://doi.org/10.5281/zenodo.19911967

Reinforcement Learning–Enhanced Hybrid CAC–CR-LBT Framework For 5G/Wi-Fi 6 Coexistence

Authors: Dr. S. Vijayakumar, C. Prabhu

Abstract: The coexistence of 5G New Radio (NR) and Wi-Fi 6 in unlicensed spectrum presents significant challenges due to contention-based medium access, cross-technology interference, and heterogeneous quality-of-service (QoS) requirements. Conventional call admission control (CAC) mechanisms reduce congestion and blocking probability, but their reliance on static admission thresholds limits adaptability under dynamic traffic and channel conditions. This paper proposes a reinforcement learning (RL) enhanced hybrid CAC integrated with a collision-resolution listen-before-talk (CR-LBT) mechanism to enable intelligent and adaptive spectrum sharing between 5G NR-U and Wi-Fi 6 systems. The RL agent dynamically tunes admission thresholds based on observed network state parameters, including offered load, signal-to-noise ratio (SNR), and service demand, while the CR-LBT mechanism mitigates contention-induced collisions during channel access. A system-level MATLAB simulation model evaluates the proposed framework under varying traffic loads. Simulation results demonstrate that the proposed scheme achieves 20–25% higher throughput and approximately 15% increased system capacity compared with conventional CAC, while significantly reducing bit error rate (BER). In addition, the framework maintains ultra-low end-to-end latency below 1 ms and near-zero blocking probability for delay-sensitive services such as VoIP and real-time video. These results confirm that the proposed approach provides a scalable and intelligent solution for next-generation heterogeneous wireless networks operating in unlicensed spectrum.

DOI: https://doi.org/10.5281/zenodo.19913041

AI-Powered Door Delivery Model For Strengthening The Public Distribution System

Authors: Suresh G. A/P, Rithika S, Tejas M U, Kathir Kamesh K, Santhosh R

Abstract: The Public Distribution System (PDS) plays a vital role in ensuring food security by distributing subsidized food commodities to economically weaker sections of society. However, the traditional distribution model faces several operational challenges including long queues at ration shops, lack of transparency, manual record keeping, and leakage of commodities during distribution. These limitations reduce the efficiency and reliability of the system. This paper proposes an AI-Powered Trusted Community Delivery Model designed to modernize the Public Distribution System using intelligent monitoring and digital verification mechanisms. The proposed system integrates artificial intelligence techniques such as route optimization algorithms, GPS-based tracking, OTP-based delivery authentication, and centralized administrative dashboards. The model introduces verified community delivery personnel who distribute ration commodities directly to beneficiaries at their homes. The system ensures transparency and accountability through real-time monitoring and automated delivery confirmation. The proposed framework has been validated through a prototype implementation and simulation-based analysis to demonstrate its feasibility and effectiveness. The results indicate that the model can serve as a practical and scalable solution for enhancing public welfare distribution systems.

Virtual Campus Assistant (VCA): A Centralized Web-Based Academic Management System

Authors: Mrs. K. Parameswari, Haritha H, Ajith Kumar B, Abinesh M, Gowtham D

Abstract: The Virtual Campus Assistant (VCA) is a comprehensive web-based academic communication and management system designed to streamline interactions among administrators, faculty members, and students within an educational institution. Conventional academic communication systems rely heavily on emails, notice boards, spreadsheets, and manually maintained records, which often leads to redundancy, delays, lack of transparency, and increased administrative workload. The proposed system addresses these challenges by integrating all academic and communication activities into a centralized digital platform. The system offers role-based dashboards for administrators, faculty, and students, a PDF-based Academic Assistant, and JWT-secured authentication. It is developed using React, Node.js with Express, and Firebase Firestore. Performance evaluations confirm improved response times, streamlined workflows, and enhanced academic transparency. The VCA provides a scalable foundation for smart campus ecosystems.

DOI: https://doi.org/10.5281/zenodo.19919777

Modern Ayurveda Platform with IoT-Based Health Monitoring

Authors: Ms. Neha Bhalerao, Ms. Pallavi Satpute, Ms. Purva Patil, Mrs. Sarojini Naik

Abstract: This work proposes the design and implementation of an Internet of Things (IoT)-based health monitoring system integrated with Ayurvedic healthcare principles. The system utilizes an ESP32 microcontroller and non-invasive sensors to measure kay physiological parameters such as heart rate, blood oxygen saturation (SpO2), body temperature, and pulse signals. The collected data are processed in real time and displayed on a web-based dashboard for continuous monitoring. The proposed system supports the Ayurvedic approach of preventive and holistic healthcare by enabling continuous observation of vital parameters and early identification of imbalances in the body. Unlike conventional monitoring systems, the platform is designed to assist in maintaining overall well-being rather than only detecting diseases. The architecture emphasizes low cost, portability, and accessibility, making it suitable for home healthcare and remote applications. By combining modern IoT technology with traditional Ayurvedic concepts, the system provides a comprehensive framework for proactive health management and improved quality of life

DOI: http://doi.org/

Integrated use of BIM Tools for Structural Retrofitting: Design Development and Clash Detection to Project Management

Authors: Assistant Professor Amarsinh B. Landage, Research Scholar Faeeza I. Sayyad, Research Scholar Sonal S. Gore, Research Scholar Shweta T. Warbhe

Abstract: This study presents a comprehensive framework for applying Building Information Modelling (BIM) tools to retrofit existing structures with improved accuracy and efficiency. BIM platforms such as Autodesk Revit are used to develop precise 3D as-built models that capture geometric, structural, and material details of existing buildings. By incorporating point-cloud data and field measurements, these models provide a reliable digital representation of actual site conditions, reducing uncertainties and enhancing visualization for informed decision-making during retrofit planning. This paper present, the focus was on developing accurate BIM-based models and performing structural analysis and optimization. Tools such as Autodesk Revit 2025, SketchUp, and STAAD Pro were integrated to enable detailed modelling, simulation, and performance evaluation. The as- built BIM models served as a foundation for structural assessment in STAAD Pro, facilitating the identification of deficiencies and the selection of appropriate strengthening strategies. This integrated workflow ensured safe, coordinated, and cost-effective retrofit design while improving overall project efficiency. The outcomes of this phase demonstrate the effectiveness of BIM in supporting retrofit design through accurate modelling and structural analysis. Focus on clash detection, design coordination, and synchronized model updates to further enhance project integration. Overall, the proposed BIM-based workflow establishes a structured approach to retrofitting that integrates 3D modelling, analysis, and cost/schedule management, contributing to reduced errors and more efficient, data-driven delivery in civil and infrastructure engineering.

DOI: https://doi.org/10.5281/zenodo.20303378

A Hybrid Approach For Secure Audio Communication Using Encryption And Watermarking

Authors: Ashish Mishra, Dheeraj Chillar

Abstract: With the rapid growth of digital communication systems, ensuring the security and integrity of audio data has become increasingly important. This paper presents a hybrid approach for secure audio communication that integrates encryption and watermarking techniques to provide dual-layer protection. In the proposed system, the audio signal is first processed using transform-based watermarking methods, such as Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD), to embed hidden information for authentication and ownership verification. The watermarked audio is then encrypted using the Advanced Encryption Standard (AES) to ensure confidentiality during transmission. At the receiver side, the encrypted signal is decrypted to recover the watermarked audio, followed by watermark extraction and verification to ensure data integrity. The performance of the proposed system is evaluated using quantitative metrics, including Signal-to-Noise Ratio (SNR), Mean Squared Error (MSE), and execution time. Experimental results demonstrate that the DWT-based approach provides better audio quality and robustness compared to the SVD-based method, while maintaining acceptable computational efficiency. The proposed hybrid framework effectively combines the strengths of encryption and watermarking, offering a secure, reliable, and practical solution for audio communication systems.

Design Of A Fire-Fighting Robot For Industrial Applications

Authors: Yashraj Nayak, Om Prakash Sondhiya

Abstract: The escalating frequency and severity of fire incidents in industrial environments—spanning petrochemical refineries, warehouses, power generation facilities, and chemical manufacturing plants—demands innovative solutions that minimise human exposure to life-threatening hazards. This paper presents the systematic design, development, and experimental validation of an autonomous fire-fighting robot engineered specifically for industrial deployment. The proposed platform integrates a thermally insulated omnidirectional Mecanum-wheel chassis with an embedded multi-sensor array comprising a FLIR Lepton 3.5 infrared thermal imager, Hamamatsu UV flame sensors, a Velodyne VLP-16 three-dimensional LiDAR, and electrochemical gas detectors. A lightweight convolutional neural network, FireDetNet-v2, trained on 45,000 annotated industrial fire images, achieves a mean average precision (mAP@0.5) of 97.6% at 30 frames per second on an NVIDIA Jetson Orin NX compute module. A 150-litre onboard water–AFFF suppression module delivers agent at up to 12 bar through a two-degree-of-freedom pan-tilt nozzle gimbal, achieving a maximum throw range of 15 m. Simultaneous localisation and mapping (SLAM) via the Cartographer framework on VLP-16 LiDAR data enables autonomous navigation in GPS-denied, smoke-filled corridors. Fifty controlled fire trials spanning Classes A, B, and C across a purpose-built industrial mock facility yielded a 94% overall suppression success rate, with a mean detection latency of 5.2 s and a mean time-to-extinguishment of 41.7 s. The system satisfies IEC 61508 SIL-2 functional-safety requirements for the suppression interlock, IP67 environmental protection, and a minimum 50-minute mission endurance.

DOI: https://doi.org/10.5281/zenodo.20053206

Sentiment Analyzer: A Multi-Method Sentiment Analysis System

Authors: Anshu Verma, Anup Kumar Choudhary, Jyotiraditya Kathua, Mohit Sharma

Abstract: Sentiment analysis has emerged as a critical application of natural language processing (NLP) in the digita l age. This paper presents Sentiment Analyzer, a comprehensive multi-method sentiment analysis system that combines lexicon-based methods (VADER, TextBlob), machine learning (ML) classifiers, and ensemble techniques to provide accurate and robust sentiment detection. The system implements a modular architecture with components for text preprocessing, sentiment analysis, emotion detection, emoji analysis, and result visualization. A FastAPI-based REST API enables programmatic access, while an interactive Streamlit dashboard provides a user-friendly interface. The ML pipeline employs TF-IDF vectorization with Logistic Regression, Naive Bayes, and Support Vector Machine classifiers. Experimental evaluation on the SST-2 benchmark demonstrates ensemble classification accuracy of 91.3%, outperforming standalone VADER (71.3%) and basic Logistic Regression (81.2%). API endpoints respond in under 50 ms for single-text analysis, and batch processing of 100 texts completes in under 450 ms.

DOI: https://doi.org/10.5281/zenodo.19940790

HEART DISEASE PREDICTION (XGBoost Random Forest , And KNN )

Authors: Riya Jaiswal, Simran Sahu, Prince Pandey

Abstract: Heart complaint remains one of the leading causes of mortality worldwide, making early discovery pivotal for effective treatment and forestallment. This design focuses on developing a prophetic model to identify the threat of heart complaint in individualities using machine literacy ways. By assaying patient data, including vital health pointers similar as age, blood pressure, cholesterol situations, casket pain type, and other applicable medical attributes, the model aims to classify individualities grounded on their liability of developing heart complaint. colorful bracket algorithms are applied and compared to determine the most accurate approach. The results demonstrate that machine literacy can serve as a dependable tool for aiding healthcare professionals in early opinion, enabling timely intervention, and eventually perfecting patient issues.

Role of Discrete Mathematics in Artificial Intelligence

Authors: Mr.Ghadage V.D.

Abstract: Artificial Intelligence (AI) has become one of the most transformative technologies of the modern era. Behind intelligent systems such as expert systems, recommendation engines, robotics, and autonomous vehicles lies a strong mathematical foundation. Among these foundations, Discrete Mathematics plays a significant role in designing algorithms, logical reasoning, knowledge representation, and decision-making processes. Concepts such as graph theory, logic, set theory, combinatory, Boolean algebra, and probability form the backbone of many AI techniques. This paper explores the importance of Discrete Mathematics in AI, its major applications, advantages, and future scope. The study highlights how discrete structures improve computational efficiency, optimize problem-solving, and support intelligent decision systems. Recent literature also emphasizes graph theory, combinatory, formal languages, and optimization methods as key foundations for AI and machine learning.

DOI: http://doi.org/

Design And Development Of An Owasp Immersive Web Security Lab

Authors: Sasmita.M, Sangeetha.V, Dr. Kamalakkannan S

Abstract: With the rapid growth of web applications, security vulnerabilities have become a major concern in modern systems. Many applications remain vulnerable due to improper input validation and insecure coding practices. This project presents an OWASP Immersive Web Security Lab designed to provide hands-on learning of common web vulnerabilities such as SQL Injection (SQLi), Cross-Site Scripting (XSS), Command Injection, File Upload Vulnerabilities, and Path Traversal. The system uses a controlled environment where users can perform attacks and observe system behavior in real time. The platform is developed using web technologies and Docker-based containerization to ensure safe and isolated lab environments. It also includes quizzes and automated verification to evaluate user performance. The results show that the system improves understanding of web security concepts and helps users learn secure coding practices effectively. This project serves as a practical educational tool and a foundation for advanced cybersecurity training systems.

DOI:

Obstacle Detection and Warning System for Hill Roads to Minimize Accidents

Authors: Muskan Nasir Dange, Sayli Jaywant Yadav, Rohit Arjun Jamdade, Sakshi Sharad Sawant, Shraddha Nitin Jamdade, Sangram Bajrang Koli, Mr. Kuldeep B. Pawar

Abstract: This paper presents a safety-oriented system designed specifically for accident prevention in ghat (hilly) road sections. The primary objective of the proposed system is to reduce accidents caused by limited visibility on sharp curves and narrow roads, where drivers are unable to detect oncoming vehicles from the opposite direction. High vehicle speed and lack of real-time information further increase the risk of collisions in such regions. To address these challenges, an embedded system based on the ATmega328 (or 8051) microcontroller is proposed. The system integrates multiple sensors, including IR or ultrasonic sensors for vehicle detection and a piezoelectric sensor for landslide detection. Additional components such as LED indicators, an LCD display, a GSM module, and a buzzer are incorporated for alert generation and communication. When a vehicle is detected at one end of the road, the system activates a warning signal using a flashing red light and alerts drivers, while a green signal indicates that the road is clear. In the event of a landslide, the system automatically triggers an early warning mechanism and closes safety gates to prevent vehicle movement, thereby reducing the risk of accidents. The proposed solution aims to enhance road safety and provide a reliable, real-time alert system for hilly terrains.

DOI: http://doi.org/

Smart HR Systems: Integrating Blockchain And AI For Transparent Recruitment

Authors: Dr. G Manjula, Kamini Sunil Bhardwaj

Abstract: Combining AI and blockchain in one solution will have a revolutionary impact on overcoming three main problems with traditional recruitment practices including inefficiency, transparency issues in decisions, and false credentials. This paper outlines an advanced concept for Smart HR Systems that is based on AI candidate screening and blockchain credential verification. The design uses NLP and deep learning techniques for resume parsing to deliver a 97.58% screening accuracy, whereas the Ethereum blockchain technology is used to secure immutable data about the verified credentials. The system utilizes the dual-approval approach that requires digital signatures by qualified experts for vacancy validation. Results obtained in six test scenarios confirm an 82% matching accuracy between the candidates’ and roles’, more than 1000 transactions per second throughput, and over 99% improvements in the performance metrics related to credential verification.

DOI: https://doi.org/10.5281/zenodo.19949274

Pragmatics Of Human–AI Dialogue: A Socio-Linguistic Study Of Conversational Agents

Authors: DR. Anjali, Rajkumar S

Abstract: While conversational agents have become increasingly ubiquitous, there remain pragmatic differences between human interaction and AI systems. This paper conducts a socio-linguistic study of the pragmatic operations involved in the language production of conversational agents with regards to speech acts, Gricean maxims, common ground, and repairs. Through the use of a dataset comprising of 120 conversations made through artificial intelligence chatbots as well as 120 human-to-human conversations, it becomes clear that conversational agents rely heavily on representative speech acts and directives but overlook expressive and commissive speech acts to generate lifeless, robotic conversations. Examples of Gricean maxim breaches involve quality (lying), relation (irrelevant information), manner (ambiguous expressions), and quantity (unnecessary elaboration). The concept of the computer as a social actor is responsible for humans perceiving conversational agents as social actors.

DOI: https://doi.org/10.5281/zenodo.19949359

Deepfake Detection Using EfficientNet-Based CNN with Threshold Optimization

Authors: Tauqeer Alam, Shashank Verma, Roshaan Raza Khan, Suman Devi

Abstract: The rapid advancement of deep learning has significantly improved the ability to generate realistic synthetic media, commonly referred to as deepfakes. While such technologies offer benefits in areas such as entertainment and media production, they also pose serious risks including misinformation, identity theft, and digital fraud. Detecting deepfake content has therefore become a critical research problem. This paper proposes a deepfake detection framework based on an EfficientNet-based Convolutional Neural Network (CNN) trained on a labeled dataset of real and fake facial images. The trained model is extended to video-level detection through frame-based inference and aggregation. In addition, a threshold optimization strategy is introduced to improve classification performance by balancing precision and recall. Experimental results demonstrate that the proposed model achieves an accuracy of 87%, precision of 97%, recall of 76%, and an AUC score of 0.96. The experimental findings confirm that threshold optimization significantly improves the balance between precision and recall, enhancing the robustness of deepfake detection systems.

DOI: https://zenodo.org/records/19949456

Skill Demand Forecasting and Salary Prediction: A Multi-Granularity Analysis Using XGBoost

Authors: Md Zahidul Islam Sany, Wubo Zhang

Abstract: The rapid evolution of the labour market makes it difficult for job seekers, employers, and policymakers to anticipate which skills will be in demand and what salaries to expect. Traditional forecasting methods often fail when faced with large-scale, sparse, and non-linear job advertisement data. In this paper, we address two interconnected problems: forecasting monthly skill demand at multiple granularities (company, region, and occupation levels) and predicting salaries from job attributes. Using real job postings collected between 2021 and 2023, we construct a comprehensive dataset containing millions of rows of skill demand time series. We apply XGBoost with carefully engineered features – 12 lagged values, a rolling 3-month average, and month indicators – to predict future demand. Because many months have zero demand, we evaluate performance separately on non-zero months. Our model achieves a Symmetric Mean Absolute Percentage Error (SMAPE) of 10.01% on active demand, demonstrating excellent predictive accuracy when a skill is actually needed. For salary prediction, we use job titles, locations, experience levels, and vacancy volume, obtaining an R² of 0.164 – modest but better than a baseline mean prediction. Beyond forecasting, we provide feature importance analysis (the rolling average is the strongest predictor), granularity comparisons (occupation-level forecasts are most accurate), clustering of jobs into four distinct market segments, and correlation analysis (experience correlates most strongly with salary). All code and processed data are publicly available to ensure full reproducibility

DOI: https://zenodo.org/records/19950109

Hostel Management System

Authors: Balachandar R, Kathir M, Suryaprakash, Ravi M, Shiyamsundar

Abstract: The Hostel Management System (HMS) is a comprehensive web-based platform developed to modernize and automate the administrative operations of educational institution hostels. Conventional hostel management relies on paper-based records, manual fee collection, and physical notice boards, leading to delays, errors, and communication gaps. The proposed system integrates role-based dashboards for administrators and students, enabling seamless management of fee collection, announcements, gate pass requests, complaints, QR-code-based attendance, and room allocation. The system is developed using React.js, HTML5, and CSS3 for the frontend, Node.js with Express.js as the backend API layer, and MongoDB as the database. Performance evaluations confirm improved operational efficiency, real-time data access, and high user satisfaction. The HMS provides a scalable and secure foundation for smart hostel ecosystems.

DOI:

Battery Management And Charging System For Solar Energy Storage

Authors: P.Prakash, P.Pushparani, Dr.M.Malarvizhi

Abstract: The utilization of BESSs is crucial for handling the issue of intermittency in the production of solar photovoltaic (PV) energy in order to provide an effective means of energy distribution in domestic and industrial sectors. This work reviews various developments made in the field of batteries and charging mechanisms of energy storage devices used for harnessing solar energy. Developments related to DC-DC converter designs, MPPT strategies, algorithms for charging and state estimations have been critically analyzed. The study reveals that the advanced versions of MPPT algorithms such as cuckoo search and incremental conductance provide high tracking efficiencies in the range of over 98%. Furthermore, physics-informed neural networks provide a better approach towards estimating SoC and SoH parameters in partial cycling conditions. It has been experimentally proved that a dual-axis tracking system along with MPPT improves average power by 71.73% when compared with fixed systems. An efficiency of 98.93% has been achieved in energy conversion.

DOI: https://doi.org/10.5281/zenodo.19958464

A Hybrid Approach For Secure Audio Communication Using Encryption And Watermarking

Authors: Ashish Mishra, Dheeraj Chillar

Abstract: With the rapid growth of digital communication systems, ensuring the security and integrity of audio data has become increasingly important. This paper presents a hybrid approach for secure audio communication that integrates encryption and watermarking techniques to provide dual-layer protection. In the proposed system, the audio signal is first processed using transform-based watermarking methods, such as Discrete Wavelet Transform (DWT) and Singular Value Decomposition (SVD), to embed hidden information for authentication and ownership verification. The watermarked audio is then encrypted using the Advanced Encryption Standard (AES) to ensure confidentiality during transmission. At the receiver side, the encrypted signal is decrypted to recover the watermarked audio, followed by watermark extraction and verification to ensure data integrity. The performance of the proposed system is evaluated using quantitative metrics, including Signal-to-Noise Ratio (SNR), Mean Squared Error (MSE), and execution time. Experimental results demonstrate that the DWT-based approach provides better audio quality and robustness compared to the SVD-based method, while maintaining acceptable computational efficiency. The proposed hybrid framework effectively combines the strengths of encryption and watermarking, offering a secure, reliable, and practical solution for audio communication systems.

Smart Railway Crossing Safety System Using Ultrasonic Sensor And Gsm

Authors: Vedikola Amarnadh Reddy , Lakshmanan.S

Abstract: Railway level crossings remain one of the most critical accident-prone zones in transportation infrastructure worldwide. This paper presents the design and implementation of an IoT-based Smart Railway Level Crossing Safety System that integrates ultrasonic obstacle detection, GPS-based train tracking, GSM-based SMS alerting, and servo motor-controlled automated gate management. The proposed system continuously monitors the crossing gate area using an ultrasonic sensor. When a vehicle becomes stranded inside the gate, the system detects the obstacle and triggers a sequence of intelligent responses: automatic gate opening via a servo motor to allow the vehicle to escape, real-time SMS alerts transmitted to the locomotive driver’s mobile device, and visual warnings displayed on the train engine control panel. Furthermore, GPS coordinates of the approaching train are tracked at predefined thresholds — at 3 km, the gate opens proactively; at 2 km and 1 km, escalating alerts are issued. Simulation and hardware prototype testing confirmed a system response time of under 1.5 seconds for gate actuation and SMS delivery. The proposed system significantly reduces the risk of train-vehicle collisions at unmanned and semi-manned level crossings.

DOI:

A Hybrid AI And User-Driven Approach For Prompt Evaluation And Ranking In Large Language Models

Authors: Mrs. Pratiksha Shevatekar, Mrs. Pooja Mishra, Prasad Patel, Raghav Tandulkar, Lokesh Dhoble, Sairaj Patil

Abstract: Unexpected progress in smart programs now shapes daily work in writing, programming, code fixes. As big language tools appear, old ways shift fast – yet better results do not come automatically. Often, flaws stem less from tech limits than weak human requests. Unclear questions tend to produce mismatched answers, raising workload while slowing output. While some sites allow sharing those queries, few offer solid ways to judge their strength, leaving users sifting through clutter where most entries add little value. Facing these issues, PromptX applies a structured method meant to improve how prompts work in practice. Through automatic assessments paired with insights from users, it judges what truly performs well. Instead of favoring only widely used options, relevance within context takes priority. As selections adapt gradually, less useful repetitions fade out naturally. Suggestions shift quietly based on individual habits and changing needs. Performance stays stable whether crafting text, solving problems, or generating code. With repeated use, refinements emerge without interruption. The system evolves simply because it is used

DOI: https://doi.org/10.5281/zenodo.19960004

VIGILFY Real Time Exam Surveillance Tool

Authors: Seetha Lakshmi R, Vijaya harshitha R, Saranya P

Abstract: In this project, it has been proposed the use of AI and ML for the purposes of real-time surveillance and detection of malpractices in an online examination system. As there could be many ways that a student may try to commit unfair practices, such as utilizing external aid, multiple candidates, or any other suspicious browsing activity; therefore, the current methodologies do not guarantee the absolute correctness and integrity of an examination. Video analysis using AI and monitoring browser activity have been combined to analyze continuous flows of information in order to detect anomalies like non-attendance, multiple people being in view, and distracted attention. The identified violations are marked and saved, and then the corresponding clips are selected to create reports that help the educators review the candidate’s actions during the test. It will help minimize the necessity for manual control and will ensure higher precision and consistency. The purpose of the project is to enable real-time detection and analysis of any exam activities by means of monitoring with intelligence.

DOI: https://doi.org/10.5281/zenodo.19960772

RecallX: An Intelligent AI-Based Multimodal Memory Recording and Retrieval System

Authors: Satyam Kumar Tripathi, Raj Singh, Jyoti Maddheshiya, Ritika Singh, Mr. Sanjeev Pathak

Abstract: People today face an overwhelming volume of daily information, which often causes cognitive overload and makes it hard to manage memories effectively. Standard tools like basic note-taking apps or digital reminders usually fall short because they lack contextual awareness and have limited search capabilities. To address this, we introduce RecallX, an AI-driven memory assistant built to capture, organize, and retrieve various types of data—including text, audio, images, and video. By combining Natural Language Processing (NLP), computer vision, and speech recognition, RecallX moves past basic file storage. Instead, it structures memories with contextual awareness and creates associative links, similar to how human memory works. The system pulls out key details, tags them with metadata like timeframes and related entities, and allows users to search their memories naturally using everyday language. By relying on semantic search rather than strict keyword matching, RecallX delivers much more accurate and relevant results. Ultimately, this scalable architecture is built to lighten the user’s mental workload and boost daily productivity.

DOI: https://doi.org/10.5281/zenodo.19969977

Multi-Tenant Serverless Query Engine With Dynamic Pricing And QoS Guarantee

Authors: Nikita Datke, Vinita Shrivastava

Abstract: Serverless computing has transformed cloud application deployment by abstracting infrastructure management while providing on-demand scalability. However, current serverless query engines often lack effective mechanisms for handling multi-tenant workloads with varying Quality of Service (QoS) expectations. In such environments, rigid pricing and static resource allocation lead to inefficiencies, SLA violations, and unfair resource distribution. This paper presents a Multi-Tenant Serverless Query Engine (MTSQE) that integrates dynamic pricing models and adaptive QoS guarantees for fair and efficient resource utilization across diverse tenants. The proposed architecture employs workload profiling, priority-based scheduling, and real-time performance feedback loops to dynamically adjust pricing and execution parameters according to SLA tiers. Through simulation using CloudSim 7G and Kubernetes-based deployment tests, the system demonstrates up to 22% improvement in resource utilization, 18% reduction in SLA violations, and 16% higher cost fairness index compared to baseline serverless frameworks. These results validate the feasibility of integrating cost-aware query management with SLA-driven adaptability for next-generation multi-tenant cloud services.

DOI: https://doi.org/10.5281/zenodo.19970491

Legal Buddy: An AI-Powered Courtroom Simulation And Legal Analysis Platform Using Retrieval-Augmented Generation

Authors: Satuluri Rajeev Varma, Dr. Kamlesh Tiwari

Abstract: Access to quality legal guidance remains out of reach for a large portion of the population, particularly in developing nations where the lawyer-to-citizen ratio is unfavorably low. This paper presents Legal Buddy, a comprehensive AI-powered legal analysis and adversarial courtroom simulation platform built to bridge that gap. The system is engineered around a Retrieval-Augmented Generation (RAG) pipeline that dynamically pro-cesses user-uploaded legal documents, embeds them semantically using Google’s text-embedding-004 model, and grounds all AI-generated outputs in verified, document specific evidence rather than unconstrained model knowledge. A distinguishing feature is the stateful Adversarial Mock Courtroom Engine, which simulates live trial proceedings by instantiating dual AI roles — an Opposing Counsel and a scrutinising Judge — thereby com-pelling users to construct and defend legally coherent arguments under realistic judicial pressure. The backend is built on Python FastAPI with MongoDB for persistent vector storage, enabling full multi-tenant session isolation. All generative outputs from the integrated Google Gemini 1.5 Pro model are constrained through rigorous prompt conditioning to the IRAC (Issue, Rule, Applica-tion, Conclusion) analytical framework and Indian legal doctrine. Empirical evaluation confirms that the combination of seman-tic vector retrieval, heuristic domain classification, and IRAC conditioned generation substantially reduces hallucination rates compared to unconstrained LLM baselines. Query classification accuracy reaches 90.5accurate, and usability testing shows 91and extensible framework for deploying AI in legally sensitive, high-accountability environments while remaining accessible without institutional infrastructure. Index Terms — Retrieval-Augmented Generation, Large Language Models, Mock Courtroom Simula-tion, IRAC Framework, MongoDB Vector Storage, Indian Legal NLP, FastAPI, Cosine Similarity, LegalBERT, Named Entity Recognition.

DOI:

Design And Implementation Of Intelligent IoT-Based Environmental Monitoring System

Authors: Dr. Shubhalaxmi Mohapatra, Ch. Prasanjit Nanda, Satyadeep Das

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.19975940

Blockchain In Digital Forensics For Evidence Integrity

Authors: Prachi P. Pophale, Dr.Harsha.R. Vyawahare

Abstract: The integrity of digital evidence is fundamental to the credibility and admissibility of forensic investigations in the digital age. Traditional evidence management techniques, primarily relying on cryptographic hashes, chain-of-custody documentation, and manual procedures, are increasingly vulnerable to tampering, human error, and sophisticated cyber threats. The advent of blockchain technology offers a transformative solution by providing a decentralized, immutable, and transparent ledger system that can significantly enhance evidence integrity assurance. This paper explores the integration of blockchain into digital forensic workflows, emphasizing its potential to establish tamper-proof, verifiable, and auditable records of digital evidence throughout its lifecycle. We examine the core components of blockchain technology—including distributed ledgers, cryptographic hashing, consensus mechanisms, and smart contracts—and their applicability to forensic evidence management. The proposed blockchain-based architecture encompasses evidence collection, secure storage, access control, and verification modules, creating a robust framework that ensures the authenticity and unaltered state of evidence. We review existing platforms and protocols, such as Factom, Evidence-Chain, Ethereum, and Hyperledger Fabric, assessing their suitability and limitations within forensic contexts. Moreover, mechanisms like timestamping, digital signatures, multi-party verification, and smart contracts are discussed as means to enforce integrity, facilitate transparent chain-of-custody, and automate validation processes. Despite its promising advantages, blockchain implementation faces challenges including scalability, privacy concerns, legal compliance, interoperability, and technical complexity. The paper also highlights future research directions—such as hybrid storage models, privacy-preserving protocols, AI integration, and quantum-resistant algorithms—that aim to address these limitations. Ultimately, this study demonstrates that blockchain technology holds substantial potential to revolutionize digital evidence management, ensuring higher levels of trust, transparency, and integrity in forensic investigations, and calls for collaborative efforts to develop standardized, scalable, and legally compliant solutions for widespread adoption.

DOI: https://doi.org/10.5281/zenodo.19977399

Customer Churn Prediction Using Machine Learning Techniques

Authors: Ashok Kumar Verma, Krishna, Amar Kumar Yadav, Ayush Chaurasiya, Sanjeev Kumar Pathak

Abstract: Customer churn is one of the major challenges faced by organizations, especially in competitive industries such as telecommunications, banking, and e-commerce. Predicting customer churn helps companies take proactive steps to retain valuable customers. This research focuses on predicting customer churn using machine learning techniques including Logistic Regression, Decision Tree, Random Forest, and XGBoost. The model is trained and evaluated using publicly available datasets. Experimental results show that ensemble-based approaches like Random Forest and XGBoost outperform traditional algorithms, achieving higher accuracy and better recall rates.

Comprehensive ML Framework For Evaluating Demographic Impacts On Healthcare Access

Authors: Tanish Aggarwal, Sunil.K.Singh, Tamanna Aggarwal, Amit Chhabra, Jagmohan Aggarwal

Abstract: Patient demographics, including gender and age, play an important role in determining access to health services and options. This study examines these demographic differences across health care systems, revealing significant differences in patient engagement and service use. With the rise of machine learning, the optimization of telemedicine has emerged as a promising strategy to improve patient care. This paper explores the use of advanced machine learning algorithms to improve telemedicine by enhancing predictive capabilities and patient engagement. Patient behavior and need patterns can be identified through predictive analytics, leading to optimized scheduling and resource allocation, increasing telemedicine utilization In- corporating machine learning into telemedicine processes does not seem to provide not only improves patient communication but also provides higher quality healthcare -emphasizes memory efficiency . By exploring these innovations, the study highlights the trans-formative potential of machine learning in telemedicine, paving the way for future advancements in digital health through increased accessibility, predictive analytics, automated reminders and data-driven insights, ultimately contributing to better patient outcomes.

DOI: https://doi.org/10.5281/zenodo.20083908

AAA Implementation With Device Hardening

Authors: P.Jahnavi, M.Hemalatha, Y.Amrutha, Mrs.Sivaselvi.k

Abstract: Current perimeter-based security solutions are no longer sufficient, as organizational networks evolve rapidly due to cloud usage, remote access, and sophisticated cyber attacks. This article describes a secure network architecture that combines authentication, authorization, and accounting (AAA) with device hardening strategies that are consistent with contemporary ZeroTrustconcepts.Theproposed system isolates network traffic via VLAN segmentation, and inter-VLAN routing via a centralized router enforces regulated communication. A RADIUS server is used for centralized authentication, guaranteeing that only confirmed users have accesstonetworkresources,whileauthorizationpoliciesand accounting systems limit access and monitor activities. The report also emphasizes the efficacy of layered security, which combines AAA with device hardening techniques including secure remote access, service limits, and tight password enforcement. When compared to existing methodologies, Zero Trust Architecture (ZTA) results in considerableimprovementsinthreatdetection,reactiontime, and security breach reduction. The findings show that centralized access control, network segmentation, and continuous verification improve security posture while lowering operational risks and financial losses. This study presents a viable and scalable methodology for creating secure corporate networks, which includes both theoretical knowledge and real-world application. It concludes that combiningAAA, device hardening, and Zero Trustprinciplesresultsinastrongdefensemodelappropriate for modern digital environments.

DOI:

Research To Reality: An End To End AI System For Automated Transformation Of Academic Papers Into Runnable Software Prototypes

Authors: Taranjeet Singh, Monish Patil, Tejas Mungekar, Swaraj Gadre, Ms. Rupali Shinde

Abstract: The rapid growth of scientific publications has created a significant gap between theoretical research and practical implementation, as many academic papers lack accessible, reproducible software artifacts. This work presents an end-to-end artificial intelligence system that automatically transforms academic research papers into runnable software prototypes. The proposed framework leverages large language models and a multi-agent architecture to perform structured document understanding, methodological decomposition, code generation, and iterative validation. By integrating natural language processing, program synthesis, and automated debugging, the system extracts key algorithmic components, reconstructs experimental workflows, and generates executable code with minimal human intervention. Additionally, a feedback-driven refinement loop ensures improved correctness and reproducibility of the generated prototypes. Experimental evaluation on a diverse set of machine learning papers demonstrates the system’s ability to produce functional and coherent implementations, significantly reducing the time required to transition from research concepts to working software. This approach contributes toward bridging the reproducibility gap in scientific research and enabling faster innovation cycles through automated research-to-reality transformation.

AI Powered Interview Preparation And Guidance Portal

Authors: Nellore. Surendra Kumar, Muvva. Vignesh, Pamanji .Phani, Dr.k.Akila

Abstract: A combination of heterogeneity in the recruitment process among companies and rapidly evolving tech stacks has evolved interview preparation to be more complex than ever before, which means that job aspirants have to rely on multiple platforms to prepare for aptitude teams, technical questions, HR interviews and mock interviews leading to fragmented learning experience. In this paper, an AI-based interview preparation and guidance portal is implemented that provides a complete personalized learning ecosystem by amalgamating company-specific modules, technical and aptitude practice exercises, HR and behavioural training sessions with focus on mock interviews, as well as the chat bot feature. By utilizing artificial intelligence techniques, along with natural language processing, user behaviour gets analysed to derive adaptive learning road maps which strive for high learning efficiency, reduced preparation time, and improved interview results through a centralized data-driven approach.

Predicting Life Expectancy: A Machine Learning Approach For Improved Accuracy

Authors: Katta. Bhavani, B. Sai Teja Reddy, Revanth Sheelam, T.R.Girish Kumar

Abstract: The number of years people survive functions as a vital marker to assess both public health state and economic standards of a community. The prediction system evaluates life expectancy through multiple analysis of GDP per capita and healthcare spending and literacy rates and deathratesalongwithvariablesthatreflectlifestylechoices. The dataset includes many records collected from internationally respected sources to ensure trustworthiness. The relationship between life expectancy and its influencing factors becomes discernible using XGBoost alongside RF and MLR as machine learning algorithms. The assessmentof predictive ability for each model depends on Mean Absolute Error (MAE) and Root Mean Squared Error (RMSE) using training and testing sets. XGBoost achieves superior accuracy than other models due to its strong capability in processing non-linear relationships. Feature importance analysis helps medical practitioners and policymakers acquire vital data about the determinants which affect life expectancy.The research contributes to life expectancy predictive modeling while helping data-based decisions for resource planning in healthcare worldwide.

IoT Based ICU Bed Availability Monitoring System

Authors: A.Arafas Ali, Dr V. Vishwa Priya

Abstract: Hospitals today face constant pressure in managing intensive care units. Real-time tracking comes into play here – it shows exactly which beds are free at any moment. When ambulances race toward a facility, knowing open spots ahead saves vital minutes. Government agencies find this helpful too, particularly when outbreaks stretch medical systems thin. Linking devices through internet-connected sensors transforms how space gets handled across clinics and major wards. Automation slips quietly into daily operations, lifting response speed without drawing attention. Efficiency grows not by chance but through steady updates fed directly from room to dashboard. Unexpected surges meet quicker replies because data flows before requests even form. From clinics to speciality wards, tracking bed use helps better balance resources. Coordination strengthens when delays decrease and care flows more smoothly. Quality rises not by chance but through steady improvements in how things run. Sensors placed right on every ICU bed spot indicate whether someone is lying there. Once gathered, that info travels through a small computer brain before hopping onto wireless networks via an IoT link headed straight for a main hub or online space. Outcomes show up live inside a phone app – staff members, even those racing toward the building, can peek at open spots without stepping foot near the ward. Knowing exactly which beds are free helps hospitals shuffle resources more effectively, cuts down on busywork done by hand, and speeds decisions when seconds matter most. Built low-cost, designed to grow step-by-step, and fits neatly into modern medical centres aiming to run smarter.

Nirbhaya Band A Wearable Emergency Response System

Authors: Prof. Kiran Khedkar, Bhor Nikhil, Suraj Giri, Dhiraj Patil

Abstract: Women’s safety has become a major concern globally due to increasing incidents of harassment and violence. This paper presents an IoT-based smart wearable emergency safety system designed using an ESP32 microcontroller, SIM800L GSM module, NEO-6M GPS module, and VC-02 voice recognition module. The system provides dual activation modes: manual (panic button) and voice based triggering. Upon activation, the device transmits real-time location coordinates via SMS to predefined contacts. The proposed solution operates without internet connectivity, ensuring reliability in remote areas. Experimental validation shows quick response time, accurate location tracking (~2.5 m), and reliable alert delivery. The system is cost-effective, compact, and suitable for daily wearable use.

InterviewSaarthi: An AI-Powered Domain-Specific Mock Interview Preparation Platform With Adaptive Feedback

Authors: Riya Singh, Ankur Kumar Singh, Sandeep Chaurasiya, Vikas

Abstract: The problem of preparing for interviews has been a consistent problem for those students and job seekers who have no option to practice in a personalized setting. This paper discusses the development of InterviewSaarthi, an artificial intelligence-based mock interview preparation system for a variety of professional areas such as Web Development, Data Science, Data Analytics, Artificial Intelligence/Machine Learning, Cyber Security, and Software Engineering. The application allows users to select different levels of complexity (Easy, Moderate, and Hard), respond to interview questions using either text or their voice, and receive feedback based on various parameters of the answer’s quality. These parameters include content, relevance, organization, grammar, confidence, and clarity. The application is created using Python Flask server and an easy-to-use HTML/CSS/JavaScript interface, as well as the API from large language models (LLMs). An experiment conducted on 30 undergraduates showed a significant increase in interview preparation skills.

DOI: https://doi.org/10.5281/zenodo.20012756

Comparative Study Of Lexicon, Machine Learning, And Transformer-Based Models For Airline Sentiment Analysis

Authors: Ansh Jena, Sujit Kakade, Arya Kedar

Abstract: Sentiment analysis can help track passengers’ per-ceptions and improve the service offered by an airline due to the increasing importance of social media, such as Twitter. It is about conducting a comparative analysis of three models of natural language processing, namely lexicon-based, machine learning, and transformer-based classification techniques for determining sentiments of airline tweets. Twitter US Airline Sentiment was chosen to be analyzed as it comprised labeled tweets from the major U.S. airlines. Data quality was improved by applying methods of text preprocessing, such as removing noise, tokeniz-ing, and eliminating stopwords. Lexicon-based sentiment analysis relied on VADER polarity baselines, machine-learning approach entailed extraction of TF-IDF features and further application of Random Forest classification technique while transformer model applied RoBERTa to identify the context of sentiment. As a result of the analysis, it was found out that while the lexicon model was faster and provided more easily understandable results, machine-learning model allowed identifying sentiments more accurately. Transformer-based RoBERTa performed the best in terms of handling more complex linguistic structures, such as negations and sarcasm.

DOI: https://doi.org/10.5281/zenodo.19948612

Hospital Workforce Scheduling and Management System Using AI-Based Optimization

Authors: P Krishna Swamy Reddy Kaduluri, Dr. Deepak K. Sinha

Abstract: Hospital workforce scheduling is one of the most operationally critical and administratively burdensome tasks in healthcare management. Manual scheduling is error-prone, time- intensive, and frequently results in unfair shift distribution, staff burnout, and inadequate resource allocation. This paper presents the design, implementation, and evaluation of a Hospital Workforce Scheduling and Management System that leverages Generative AI — specifically OpenAI’s GPT-3.5-turbo — to automate and optimize weekly shift scheduling across multiple hospital departments. The system integrates a Flask-MongoDB backend with a dual-mode scheduling engine: an AI-primary mode that generates context-aware, department-specific shift al- locations, and a deterministic Round-Robin fallback that ensures 100% scheduling availability even when the external API is unavailable. The platform supports four distinct user roles — Doctor, Nurse, Receptionist, and Administrator — each with tailored dashboards and access controls. A comprehensive Leave Management System with automated replacement assignment from a general backup pool and email-based notifications is also implemented. Experimental results demonstrate a reduction in scheduling time of approximately 80%, schedule generation in under 5 seconds via AI, and sub-100 ms generation via the fallback algorithm. Security vulnerabilities are identified and a remediation roadmap is established. The system serves as a production-ready, cost-effective, and extensible reference imple- mentation for AI-powered healthcare workforce management.

BrewHaven: A Browser Based Cafe Ordering And Management System

Authors: Anshika Saini, Ayesha Garg, Nishant, Shivani Panchal, Ankur Kaushik

Abstract: This paper presents BrewHaven, a café ordering and management system built entirely using HTML, CSS, and JavaScript— requiring no server, no database software, and no internet connection. The system consists of two self-contained files: a customer-facing ordering interface and an admin dashboard, both communicating exclusively through the browser’s localStorage API. BrewHaven demonstrates that modern web technologies alone are sufficient to deliver a functional, visually polished digital ordering experience for small cafés. This paper describes the system architecture, design decisions, feature set, limitations, and proposed future enhancements.

DOI: https://doi.org/10.5281/zenodo.20019420

Advances In Chemical Functionalization And Engineering Of Optical Fiber Composites And Multifunctional Sensor Complexes In Robotics

Authors: Sandy Subala Soosai Michael

Abstract: Optical fibers have emerged as critical enablers in modern robotics. In particular, their integration into composite structures and functionalized sensor complexes supports real-time sensing, proprioception, and environmental interaction. This review comprehensively examines the chemical components (core and cladding materials such as silica glass and polymethyl methacrylate (PMMA)), composite materials (fiber-reinforced elastomers and carbon-fiber reinforced polymers (CFRP) with embedded optical fibers), and sensor complexes (indicator-doped coatings, sol-gel matrices, and biorecognition elements) that underpin fiber-optic technologies in robotic systems. Key applications span soft robotics (shape sensing via fiber optic shape sensors (FOSS) and polymer optical fibers (POF)), continuum manipulators, exoskeletons, and structural health monitoring (SHM) in rigid robotic platforms. Moreover, advances in stretchable waveguides, multicore fibers, and hybrid composites enable submillimeter resolution in curvature, strain, twist, collision detection, and stiffness perception. However, challenges including cross-sensitivity (strain-temperature), mechanical delamination, integration complexity, cost, and data processing overhead limit widespread adoption. Therefore, future directions emphasize AI-enhanced signal decoupling, scalable 3D-printed multifunctional fibers, and bioresorbable complexes for biomedical robotics.

DOI: https://doi.org/10.5281/zenodo.20019589

Deepfake Detection Using Cnn-Lstm: A Video Based Approach

Authors: Padala Sri Roshni, Dr. Poorna Mishra

Abstract: In recent months, free deep learning-based software tools has facilitated the creation of credible face exchanges in videos that leave few traces of manipulation, in what they are known as “DeepFake”(DF) videos. Manipulations of digital videos have been demonstrated for several decades through the good use of visual effects, recent advances in deep learning have led to a drastic increase in the realism of fake content and the accessibility in which it can be created. These so-called AI-synthesized media (popularly referred to as DF). Creating the DF using the Artificially intelligent tools are simple task. But, when it comes to detection of these DF, it is major challenge. Because training the algorithm to spot the DF is not simple. We have taken a step forward in detecting the DF using Convolutional Neural Network and Recurrent neural Network. System uses a convolutional Neural network (CNN) to extract features at the frame level. These features are used to train a recurrent neural network (RNN) which learns to classify if a video has been subject to manipulation or not and able to detect the temporal inconsistencies between frames introduced by the DF creation tools. Expected result against a large set of fake videos collected from standard data set. We show how our system can be competitive result in this task results in using a simple architecture.

DOI: https://doi.org/10.5281/zenodo.20019797

Recycling Of Wastepaper To Eco-Friendly Mosquito Repellent Sticks By Using Neem And Papaya Leaves

Authors: Asavari Bhosalea, Manish S. Warudea, Aishwarya S. Patila, Tushar A. Shindec

Abstract: Waste is the term used to describe undesired resources created from domestic, commercial, industrial, and institutional activities. It can be classified based on its origin and composition, such as organic matter, paper, glass, metals, and plastics. In there are so many educational institutions, workplaces, and packing industries generate large amounts of paper waste, contributing significantly to municipal solid waste. Paper pulp can be produced from wood, plant materials, and recycled paper. Neem (Azadirachta indica) and papaya (Carica papaya) are valuable plants having excessive medicinal and insecticidal properties. In Papaya leaf extract contains cyanogenic glycosides, papain, and alkaloids which is effective against insect or pests. Camphor, a natural tree-derived compound, is widely used in mosquito repellents due to its strong odour and has long-lasting repellent action.

DOI: https://doi.org/10.5281/zenodo.20022338

Human-Robot Collaboration In Industries

Authors: Raaj Mahadik, Dr.Asha Durafe, Kunal Gopalkar, Vismay Parab, Jeet Joshi, Harsh Bane, Avnish Kubal, Tejas Dalvi

Abstract: Human-Robot Collaboration (HRC) has emerged as a foundational paradigm of the fourth industrial revolution (Industry 4.0), enabling seamless integration of human cognitive capabilities with robotic precision, strength, and repeatability within shared workspaces. This paper presents a comprehen- sive survey of HRC in industrial environments, synthesizing recent advances across sensor-based perception, safety control techniques, collaborative robot (cobot) design, and workforce competency requirements. Drawing from an analysis of fifteen key studies spanning assembly, manufacturing, construction, and smart factories, we examine the classification of HRC in- teraction modes—coexistence, synchronization, cooperation, and collaboration—alongside critical enabling technologies including deep learning-based visual perception, motion planning, collision avoidance, ergonomics-oriented control, and digital twins. A structured literature review table, comparative analysis, and key findings discussion are presented. The paper identifies persistent challenges in data collection, model reliability, safety assurance, and human competency alignment, while highlighting promising research directions toward Industry 5.0, where human-centric collaboration becomes the central focus.

Isolation And Screening Of Pectinolytic Microbial Strains: A Comprehensive Review

Authors: Ananya Chaudhary, Almas Nauseen, Misbah Fatima Rizvi, Ashima Kathuria

Abstract: Pectinolytic enzymes, collectively known as pectinases, constitute a major class of industrial biocatalysts with extensive applications in food processing, textiles, paper and pulp, bioenergy, and waste management. These enzymes are widely employed in fruit juice clarification, wine stabilization, textile bioscouring, coffee and tea fermentation, paper pulping, and the bioconversion of agro-industrial residues into value-added products. Microorganisms serve as the most efficient and economical source of pectinases due to their rapid growth, extracellular enzyme secretion, and ease of genetic and process optimization. Recent advances emphasize sustainable and large-scale production strategies, including the use of low-cost agro-waste substrates, optimized fermentation systems, and bioreactor-based processes. This review integrates the comprehensive information on the isolation, screening, and characterization of pectinolytic microbial strains, highlighting methodological approaches, microbial diversity, enzyme assays, and industrial relevance, while providing a coherent framework for identifying strains suitable for commercial exploitation.

DOI: https://doi.org/10.5281/zenodo.20023251

Fake News Detection And Verification With Machine Learning

Authors: N. Chetanaditya, K. Likith Avinash, K. Naveen Sai, P. Teja Prasanth, K. Shiva Prasad, Dr. Venkataramana D, Dr. Rajaprakasha Rao P

Abstract: The rapid growth of digital media platforms and social networks has significantly increased the spread of misinformation and fake news, posing serious threats to public opinion, social stability, and democratic processes. This paper presents a machine learning-based approach for the detection and verification of fake news by analyzing textual content and metadata associated with news articles. The proposed system leverages Natural Language Processing (NLP) techniques for feature extraction, including tokenization, stop-word removal, and vectorization methods such as TF-IDF. Various supervised machine learning algorithms, including Logistic Regression, Support Vector Machines, and Random Forest, are employed to classify news articles as genuine or fake. To enhance verification, the system incorporates credibility scoring based on source reliability and linguistic patterns commonly associated with misinformation. A comparative performance analysis is conducted using standard evaluation metrics such as accuracy, precision, recall, and F1-score on benchmark datasets. Experimental results demonstrate that the proposed model achieves high classification accuracy while maintaining robustness against diverse writing styles and misleading patterns. The system is designed to be scalable and adaptable, enabling real-time detection and integration with social media platforms. This research highlights the potential of combining machine learning techniques with linguistic analysis to combat the growing challenge of fake news dissemination. Future work includes incorporating deep learning models and fact-checking APIs to further improve verification accuracy and system reliability.

Multi-modal Fake News Detection

Authors: Vishal Rajak, Kundan, Mayank Gautam, Mrs. Naimisha awasthi

Abstract: The rapid advancement of Artificial Intelligence (AI) has significantly enhanced the ability to generate highly realistic synthetic content, including deepfake images, videos, and misleading textual information. While these technologies offer innovative applications, they also pose serious threats in the form of misinformation and digital manipulation. Detecting such content has become increasingly complex due to the sophistication of modern AI models. This research proposes a comprehensive multimodal framework for detecting fake news and deepfake media by integrating multiple AI techniques. The system utilizes Convolutional Neural Networks (CNNs) for image analysis, frame-based processing for video deepfake detection, and transformer-based Natural Language Processing (NLP) models such as BERT for text classification. Additionally, external fact-checking APIs are incorporated to validate information in real time. To enhance interpretability, the system employs Grad-CAM visualization techniques that highlight manipulated regions within images, enabling users to better understand model decisions. The proposed approach leverages the strengths of each modality to improve detection accuracy and robustness. Experimental results demonstrate that the multi-modal system achieves superior performance compared to traditional single-modality approaches. The system is scalable, efficient, and suitable for real-world deployment in combating the spread of misinformation across digital platforms.

DOI: https://doi.org/10.5281/zenodo.20024641

AI-Based Smart Study Planner

Authors: Ankit Kumar, Sarvesh Kumar, Anup Sharma, Sakshi Rastogi

Abstract: In today’s academic environment, students face significant challenges in managing their time effectively due to multiple responsibilities such as attending classes, completing assignments, preparing for examinations, working on projects, and developing technical as well as soft skills. Poor time management often leads to stress, reduced productivity, and poor academic performance. This paper proposes a Smart Student Time Management System, which helps students efficiently organize their daily academic and personal tasks. The system integrates task scheduling, priority management, and intelligent reminders to optimize time utilization. It allows users to input their tasks, assign priorities, and receive automated suggestions for scheduling based on urgency and importance. The proposed system uses rule-based logic and optional machine learning techniques to analyze user behavior and recommend optimized study schedules. The system also tracks progress and provides feedback to improve productivity over time. Experimental evaluation shows that the system significantly enhances task completion rate and reduces procrastination among students.

DOI: https://doi.org/10.5281/zenodo.20029286

Lightweight MobileNetV2–Attention Framework For Maize Leaf Disease Classification Under Semi-Realistic Conditions

Authors: Amrita Kumari, Ankush Kumar, Omvir Singh

Abstract: Leaf diseases significantly affect the Maize productivity. Early and timely detection of these diseases help in reducing crop loss. It also ensures sustainable agricultural practices. Traditional examination of leaves by means of manual methods are protracted, subjective, and difficult to scale, especially when working on large agricultural fields. In the current work, a lightweight hybrid deep learning (DL) framework for maize leaf disease detection has been developed. The proposed framework is based on MobileNetV2 integrated with a spatial attention mechanism. The MobileNetV2 backbone allows effective feature extraction and the attention module helps in improving the model’s ability to focus on the regions affected by disease. To enhance generalization and reduce overfitting, several training strategies are employed. These include MixUp augmentation, exponential moving average (EMA) and label smoothing. The evaluation of the model is performed on datasets with semi-realistic conditions with moderate background and illumination variability. A peak accuracy of approximately 97.45% was achieved by the proposed approach as confirmed by the evaluation results and exhibits stable convergence. These findings indicate that lightweight architectures, when synergistically combined with attention mechanisms and robust training strategies, can attain high accuracy. The architecture also preserves computational efficiency. This makes them suitable for real-world agricultural deployment.

DOI: https://doi.org/10.5281/zenodo.20030190

A Modern Content Management System Front-End Interface: Design, Implementation, And Performance Analysis

Authors: Ankit Prajapati, Ashish Shukla, Prem Chandra, Ankit Jaiswal

Abstract: This paper presents the design and development of a modern Content Management System (CMS) front-end interface, addressing key challenges in traditional CMS platforms, such as poor flexibility, high technical complexity, and lack of responsive design. The proposed system integrates user authentication features, including login, signup, and password recovery, along with a centralized dashboard and content management modules for adding and viewing digital content. The system follows a component-based architecture that promotes code reusability and scalability while employing efficient rendering techniques that update only the necessary parts of the interface instead of reloading entire pages. The user interface was designed for simplicity, clarity, and responsiveness, ensuring consistent performance across desktops, tablets, and mobile devices. The system was tested across multiple user sessions and device types, demonstrating a 100 percent functionality pass rate on all core modules. Performance analysis revealed average response times below 300 ms, with user satisfaction scoring 4.7 out of 5. The results confirm that the proposed CMS front-end interface successfully reduces the complexity for non-technical users and provides a scalable foundation for future enhancements, such as backend integration and advanced security mechanisms.

DOI: https://doi.org/10.5281/zenodo.20034011

A Web-Based Stroke Risk Prediction System Using Ensemble Machine Learning: Development, Evaluation, And Clinical Utility

Authors: T. T. Visali, D. Harini, S. Prathi

Abstract: Stroke is a leading global cause of mortality and long-term disability, yet the majority of strokes are preventable through early risk stratification and timely clinical intervention. This paper presents the design, implementation, and evaluation of a web-based stroke risk prediction system that integrates ensemble machine learning with a Flask-based clinical decision support interface. Five classification algorithms — Logistic Regression, Decision Tree, K-Nearest Neighbours, Support Vector Machine (SVM), and Random Forest — are trained and compared on the publicly available Kaggle stroke prediction dataset (n = 5,110 records, 11 clinical and demographic features). Class imbalance, which afflicts 95.13% of records as non-stroke, is addressed through Synthetic Minority Over-sampling Technique (SMOTE) before model training. Random Forest achieves the highest performance, with an accuracy of 88.7%, precision of 87.4%, recall of 83.2%, F1-score of 85.3%, and AUC-ROC of 0.918. The serialised model is deployed through a Flask web application that accepts eleven clinical inputs, executes real-time inference, and returns a binary stroke risk prediction with an explanatory probability score. Comparative benchmarking against four published stroke prediction studies confirms that the proposed system achieves competitive accuracy and is the only implementation among the compared works to integrate both SMOTE-balanced ensemble modelling and a deployable web interface within a unified pipeline. The system is intended as a low-cost clinical decision-support tool for healthcare practitioners and risk-aware individuals in resource-limited settings.

DOI: https://doi.org/10.5281/zenodo.20035262

Modeling and Analysis of Wear behavior in Angular Contact Ball Bearingss

Authors: Md Kawsar Afjal Shambit, Isyaku Muhammad

Abstract: This paper presents a finite element methodology for predicting time to failure of SKF 7218 angular contact ball bearings based on wear degradation using the Archard wear model. A comprehensive 3D simulation was conducted in Ansys Workbench 2026 R-2 under aerospace-relevant operating conditions: rotational speed of 6000 rpm and axial load of 5000 N. Contact pressure and sliding distance were extracted over 30 simulation steps and utilized to compute wear depth evolution. The results demonstrate a nonlinear wear progression characterized by three distinct phases: an initial running-in period (0–10 mm sliding distance, wear depth 0–1.0 µm), a steady-state regime (10–40 mm, 1.0–2.0 µm), and a mild wear stabilization beyond 40 mm (2.0–2.4 µm). A parametric study on material hardness (1500–2500 MPa) revealed an inverse relationship with wear depth, reducing wear from 97 µm to 49 µm at 58,000 cycles. The wear depth versus contact pressure curve exhibited an accelerating trend, transitioning from mild wear (0.06 µm at 1.55 MPa) to severe adhesive/abrasive wear (61.5 µm at 53.64 MPa). These findings provide quantitative benchmarks for condition-based maintenance scheduling and pre-failure analysis in aerospace auxiliary systems including aircraft accessory gearboxes, fuel pumps, and UAV propulsion systems.

DOI: https://zenodo.org/records/20035276

In Silicon Study Of Plant Based Phenolic Compounds

Authors: Sonali Mahadev Patil, Dhanashri Shivanand Chavan, Sanika vasant salgare, Laxmi Dhundappa Narute

Abstract: Phenolic compounds (PCs) are plant-derived secondary metabolites recognized for their diverse biological activities, including antioxidant, anti-inflammatory, and anti-diabetic properties. While conventional oral hypoglycemic agents like metformin and glimepiride are widely used to manage blood glucose, they are often associated with adverse effects such as gastrointestinal distress, allergic reactions, and liver inflammation. This study utilized molecular docking to evaluate the anti-diabetic potential of 29 plant-based phenolic compounds compared to standard marketed drugs. The results demonstrate that all 29 natural compounds exhibited superior docking scores and glide energy relative to the reference standards, suggesting their potential as safer and more effective alternatives for blood glucose regulation.

Mechanical Ventilation in Underground Structures

Authors: Assistant Professor R. V. Bhalerao, Pratik R Kharat, Harshal S Padol, Anup R Deshmukh, Shubham S Dandade, Yogesh G Hissal, Chetan V Lahane, Dinesh D Chaudhari, Abhijeet Shevale, Shrihari J Jadhav

Abstract: This paper explores base isolation as an advanced structural engineering technique to mitigate earthquake effects. By decoupling structures from ground motion, isolators such as elastomeric and sliding bearings significantly reduce seismic force transmission. The study covers methodology, material properties, and comparative performance against fixed-base structures, concluding that base isolation is a vital component for resilient infrastructure in seismic zones.

DOI: http://doi.org/

A Real-Time Location-Aware Blood Bank Web Application: Architecture, Implementation, And Performance Evaluation

Authors: C. Lokesh, M. Sakthivanitha

Abstract: Timely access to compatible blood during surgical emergencies and trauma care is a life-critical requirement that conventional manual blood bank coordination frequently fails to satisfy, particularly in densely populated urban regions where demand-side unpredictability and supply-side fragmentation coexist. This paper presents the design, implementation, and quantitative evaluation of a web-based Blood Bank Application that integrates real-time geolocation tracking to dynamically connect blood donors, recipients, and healthcare institutions through a unified digital platform. The system is implemented using a three-tier architecture comprising an HTML5/JavaScript presentation layer, a Node.js RESTful application layer, and a Firebase Firestore cloud database, with the Google Maps API providing Haversine-based proximity ranking of available donors and Firebase Cloud Messaging delivering sub-3-second push notifications to matched donors. Experimental evaluation across 110 blood requests involving 212 registered donors across all eight ABO/Rh blood groups demonstrates an overall request fulfilment rate of 83.7%, a 60.9% reduction in average donor search time relative to conventional telephone-based coordination (from 18.4 minutes to 7.2 minutes), and a donor notification latency of under 2.1 seconds. Comparative benchmarking against five published blood bank systems confirms that the proposed implementation is the only system in the comparison group to provide the complete combination of web accessibility, real-time geolocation, cloud-native data management, and push notification in a single integrated platform. The system offers a scalable, cost-effective digital health infrastructure for emergency blood supply management in hospital, NGO, and community health settings.

DOI: https://doi.org/10.5281/zenodo.20037024

Smart Campus Engagement System: An Ai-Powered Unified Platform For Student Interaction, Academic Tracking, And Campus Service Management

Authors: B. Anief, S. Harish, M. Zhariyathazzes, Dr. M. Sakthivanitha

Abstract: Higher education institutions worldwide operate under a persistent structural burden: academic data, administrative workflows, and student support mechanisms are distributed across disconnected, incompatible platforms, creating information asymmetry and disengaging students from institutional life. This paper presents the Smart Campus Engagement System (SCES), a full-stack, AI-augmented digital platform that addresses this fragmentation gap through a unified, role-differentiated interface connecting students, teaching staff, and administrators. The system is grounded in a three-tier microservices-inspired architecture comprising a Next.js frontend, a Python FastAPI backend, and a PostgreSQL database deployed via Docker containerisation. A five-role RBAC model (Admin, Staff, Student, Hosteller, Day Scholar) governs access, while a LLaMA-3.3-70B language model, served via the Groq cloud inference API, provides 24/7 conversational campus support. Seven interdependent functional modules cover user management, academic tracking, real-time WebSocket-based push notifications, hostel and outpass management, complaint escalation, event lifecycle management, and student engagement analytics. Iterative design-build-test methodology was applied across four development sprints. System evaluation encompassed functional testing (8 representative test cases, 100% pass rate), performance testing (312 ms mean login latency; 94 ms WebSocket broadcast latency to 200 concurrent clients), security assessment against the OWASP Top-10 taxonomy, and a formative usability study (n = 40, overall satisfaction mean 4.5/5.0). Results demonstrate that SCES successfully resolves the identified fragmentation gap: latency benchmarks are within interactive-system thresholds, role enforcement eliminates cross-role access violations, and user satisfaction significantly exceeds the pre-deployment baseline. The platform operationalises the Smart Campus 4.0 vision and provides a replicable blueprint for institutions seeking to consolidate academic and administrative services under a single AI-enabled environment.

DOI: https://doi.org/10.5281/zenodo.20037098

Comparative Study Of Consumer-Grade And Clinical-Grade EEG Devices For Depression Detection

Authors: Nishchay Kumar, Shivank Soni

Abstract: Depression is a major global health concern, and early detection remains critical for timely intervention. Electroencephalography (EEG) provides a non-invasive means of identifying neurophysiological patterns associated with depressive disorders. However, traditional clinical-grade EEG systems are expensive, require complex setup, and are confined to laboratory environments. In contrast, low-cost consumer-grade EEG headsets—such as Muse, Emotiv, and OpenBCI—offer portability and affordability but are often criticized for limited channel count, lower sampling rates, and higher susceptibility to noise. This study presents a systematic comparative analysis of clinical- and consumer-grade EEG devices for automated depression detection. Using both public datasets and paired recordings, we evaluate signal fidelity, feature discriminability, and classification accuracy across multiple machine-learning and deep-learning models. The proposed evaluation pipeline (Figure 2) includes standardized preprocessing, artifact removal, and feature extraction methods, while Table 1 summarizes device specifications and Table 2 lists the datasets employed. Results demonstrate that, although clinical systems outperform consumer devices in signal quality and peak accuracy, optimized preprocessing and transfer-learning models significantly narrow the gap, yielding only marginal differences in classification outcomes. These findings indicate that consumer-grade EEG can serve as a viable alternative for preliminary depression screening, enabling scalable and cost-effective mental-health monitoring in real-world settings.

DOI: https://doi.org/10.5281/zenodo.20044444

Fast Google Dork Scan _369

Authors: Karan Borse, Atharv Borkar, Santosh Nimbalkar, Bhuvanesh Marathe, Prof. Sunita Parinam

Abstract: Google Dorking, also known as Google Hacking, is a technique that uses advanced search operators to discover hidden information on the web. Originally used for legitimate research and indexing, it is now a powerful tool in cybersecurity for both ethical hackers and malicious attackers. This paper explores the mechanisms, use cases, and implications of Google Dorking. It discusses how sensitive information can be exposed unintentionally, the legal boundaries surrounding its use, and preventive strategies to reduce vulnerabilities. We demonstrate the power of Google Dorking through real-world examples and emphasize the need for increased awareness among developers and administrators.

Biometric Security: Vulnerabilities And Liveness

Authors: Aditya .S. Ubhe, Prof. (Dr.) Swapnesh Taterh

Abstract: From unlocking personal smartphones to securing international borders, biometric technologies—such as facial recognition, iris scanning, and fingerprint analysis—have fundamentally transformed digital security. While these physical identifiers offer a significant upgrade over traditional passwords in both convenience and reliability, they are increasingly becoming targets for sophisticated adversaries. Attackers are continuously developing novel ways to exploit vulnerabilities, targeting both the physical sensors that capture data and the underlying machine learning algorithms that process it. This paper explores the current landscape of biometric vulnerabilities, detailing the specific tactics used to deceive these systems. Crucially, it critically evaluates current “liveness detection” protocols—the security layers built to distinguish a genuine, living user from a synthetic or spoofed input—to assess their real-world effectiveness against modern evasion techniques. Experimental simulations and comparative analyses are conducted to measure spoofing success rates, detection accuracy, and authentication error rates under multiple attack scenarios. Based on the findings, a hybrid biometric security framework is proposed that integrates multi-modal biometric verification, behavioral biometrics, and artificial intelligence–based anomaly detection. The proposed framework aims to improve resilience against both physical spoof artifacts and AI-driven adversarial attacks in modern biometric systems.

DOI: https://doi.org/10.5281/zenodo.20044732

A Comprehensive Review Of Statistical Methods In Scientific Research

Authors: Miss. Mane Mukta Nanaso

Abstract: Statistical techniques are crucial in scientific research as they facilitate the systematic gathering, structuring, analysis, interpretation, and presentation of data. These methods convert raw data into valuable insights and underpin evidence-based decision-making. This review paper outlines essential statistical methods employed in various research domains, such as descriptive statistics, which condense data, and inferential statistics, which enable conclusions about populations based on sample data. It also emphasizes significant techniques like regression analysis and hypothesis testing, commonly used to explore relationships between variables and verify research hypotheses. Additionally, the paper examines emerging strategies like machine learning and Bayesian analysis for managing complex datasets. The importance of sound research design, including sample selection and bias management, is stressed to ensure precision and dependability. In summary, statistical methods are indispensable for generating valid, reliable, and reproducible research results across different fields.

Centralized AI Process Monitoring System

Authors: Abinash Bir, Aanchal Kumari, Abhishek Rai, Bhaskar Gautam

Abstract: Now a days online education is increasing very fast and many exams are conducted in remote mode. Because of this flexibility, students can give exams from home or any place which is very convenient and useful for them. But along with this advantage, cheating methods are also increasing very fast because of AI tools, internet and different smart devices. Students can easily search answers, use AI tools, open multiple tabs, switch screens or even take help from another person sitting near them without getting noticed properly. This paper presents a smart monitoring system using artificial intelligence to reduce cheating in online exams. The system combine multiple modules like identity verification, behaviour tracking, environment monitoring and content checking. These modules continuously observe student activity and send data to centralized system where all analysis is done together. This makes system more effective and reliable compared to single method systems. The main goal of this system is to create a fair environment for all students. Honest students should not feel disadvantage because of others cheating. But there are some issues like privacy concerns, wrong detection and mental stress on students. Sometimes system can detect normal behaviour as suspicious which is not correct and may create problem for students. So it is very important to use AI system along with human judgement. This paper explain working of system, its advantages, limitations and possible improvements in future in simple and understandable way.

SyncDesk : A Real-Time Multi-User Collaboration Platform Using WebSockets

Authors: Avi Gupta, Anant Kumar, Sankalp Sharma, Yashi Singh

Abstract: SyncDesk is a multi-modal real-time collaboration platform integrating code editing, rich-text documents, interactive whiteboarding, and chat in a single web environment. It allows multiple authenticated users (e.g. students or teams) to co-edit content within a private “room.” Built with a React frontend and Node.js/Express + Socket.IO backend (MongoDB for persistence), SyncDesk propagates edits (characters or drawing strokes) word-by-word to all participants. It exploits Engine.IO transport (upgrading from HTTP polling to WebSocket) to achieve near real-time updates (observed <100 ms round-trip in local tests)[1]. The contributions of this work are (1) a unified multi-tool architecture for collaborative editing, (2) fine-grained synchronization using Socket.IO events, and (3) role-based access control with persistent versioning. These address the gap noted by Wang et al. that no existing system integrates these varied collaboration modes[2][3].

DOI:

SDG Buddy: A Holistic Framework For Individual Sustainability Tracking Using Multi-Criteria Impact Assessment And LLM-Driven Gamification

Authors: Tejas Kakani, Samyak Kankariya, Tejal Salunke, Prajval Agawane, Yashraj Patil (Guide)

Abstract: While the United Nations’ 2030 Agenda for Sustainable Development calls for universal action, contemporary sustainability applications remain predominantly focused on environmental metrics, specifically carbon footprint tracking. This paper introduces “SDG Buddy,” a novel framework designed to track, quantify, and gamify individual actions across the full spectrum of the 17 Sustainable Development Goals (SDGs). By integrating Large Language Models (LLMs) for natural language action processing and Life Cycle Assessment (LCA) principles for impact scoring, the proposed system bridges the gap between vague intentions and measurable global impact. We detail a cloud-native, serverless architecture that ensures scalability and resilience, addressing the technical and financial pitfalls of previous sustainability tracking initiatives.

Yolo Based Real Time Animal Detection And Counting System

Authors: D. D. Pukale, Vinaya Kulkarni, Riddhi Taharabadkar, Apeksha Varangane, Pranali Barge, Sakshi Sonawane

Abstract: Wildlife monitoring is an important tool for improving biodiversity conservation efforts, conducting ecological research, and managing human/wildlife interactions. Traditional methods of monitoring wildlife such as manual observations of live animals or using captured camera images usually take considerable time and resources, and therefore they are prone to human error. In this paper, we describe an array of proposed automated systems designed to enable real-time detection and identification of wild animals through captured images and video streams, utilizing deep learning techniques. For this work, we will use the YOLO (You Only Look Once) object detection algorithm in conjunction with specific computer vision techniques in order to detect and classify different animal species based on captured images and video streams. The model was trained using both annotated datasets and defined datasets containing multiple types of species (e.g., deer, elephants, tigers, leopards). The detection process includes extracting frames from a video stream and performing image preprocessing on the individual frames, applying the trained YOLO object detection model to localize (bounding box) and identify (classification) the species of the detected animal(s) from the individual frames based on a confidence score. In addition to species detection, we have also developed a multi- object tracking system that allows for consistent identification of an animal across different frames and prevents double counting of an animal. The experimental trials conducted using this system have demonstrated high detection rates and real- time performance in various environmental conditions, including motion blur, partially obscured individuals, and shadows and reflections of light. This detection system can be used to assist with the management of forest-related resources. Recommendations made by the authors include: 1) the ability to more effectively manage forest resources through increased knowledge of wildlife behavior; 2) improved training and collaborative efforts with other researchers may result in added benefits; and 3) forest managers can derive benefits from the implementation of this new technology.

Blockchain-Based Online Voting System With Vote Receipt Verification And Ai-Based Fraud Detection

Authors: J.Rajasubha, Gowthami.M, Ilamathi.T, Dinesh.S

Abstract: Secure and transparent voting mechanisms are critical for maintaining the integrity of democratic processes. Conventional voting methods and existing electronic voting systems often suffer from centralized control, limited transparency, high operational cost, and vulnerability to manipulation. To address these challenges, this paper proposes a blockchain-based online voting system that ensures decentralization, immutability, and verifiable election processes. The proposed system leverages blockchain technology to record each vote as an immutable transaction, preventing unauthorized modification and ensuring end-to-end transparency. To enhance system security, role-based access control (RBAC) is implemented to restrict operations based on predefined user roles, thereby preventing unauthorized administrative actions. Additionally, a vote receipt verification mechanism is introduced, allowing voters to independently verify the inclusion of their vote in the blockchain without compromising voter anonymity. To further strengthen election integrity, an AI-based fraud detection module analyzes voting behavior and system activity patterns to detect and mitigate suspicious activities such as multiple voting attempts, abnormal access behavior, and automated attacks. The integration of blockchain, artificial intelligence, and access control mechanisms significantly improves voter trust, election transparency, and system reliability. Experimental analysis indicates that the proposed system reduces fraud risks, enhances verification efficiency, and supports scalable online elections, making it suitable for modern digital voting environments.

LokSevaAI: Smart Governance Complaint Redressal System

Authors: Keshvee Patel, Divya Shah, Ishan Tarkas, Prof. Medha Asurlekar

Abstract: Current systems for grievance redressal in cities are predominantly manually or semi-automated based which lead to delays, misclassification and slow handling of complaints. The work proposes LokSevaAI-a smart complaint management framework which automates complaint classification, prioritization and routing based on techniques from NLP and ML. Initially the unstructured text of the complaint is pre-processed. Then features from it are extracted using Term Frequency-Inverse Document Frequency. The features are then used to classify the complaint into a specific governance area using a supervised ML model like logistic regression, support vector machine or random forest for multi-class classification. A separate module for prioritization based on the sentiment analysis of the complaint has been implemented to attend to urgent and sensitive complaints first. Based on the classification and prioritization, automated routing has been enabled. A monitoring dashboard shows real-time status of complaints and helps in analysis and decision making. Efficient management of large datasets and storing of complaint data in a structured manner has also been taken into consideration for better monitoring throughout the life cycle of a complaint. Furthermore, it supports data-driven decision-making by providing features for analyzing trends and measuring performance. The proposed solution automates most of the tasks in the existing manual system and drastically improves response time and transparency. It is computationally inexpensive and feasible for deployment on a large scale in smart cities; there is also scope for integration with multi-lingual and voice-based services.

DOI: https://doi.org/10.5281/zenodo.20050331

Krisho: Bharat’s Direct Farm-to-Table Marketplace Using AI and Voice Technology

Authors: Assistant Professor Shubhi Verma, Shaurya Tyagi, Shubh, Jhanvi Tyagi

Abstract: Krisho is an innovative digital platform designed to transform India’s agricultural supply chain by enabling a direct farm-to-table marketplace. The system addresses the long-standing issue of middlemen exploitation, where farmers receive minimal profit while consumers pay inflated prices. By eliminating intermediaries, Krisho ensures fair pricing, increased transparency, and better income opportunities for farmers across the country. The platform integrates advanced technologies such as Artificial Intelligence (AI), real-time communication, and multilingual support to make it accessible to farmers of varying literacy and technical skill levels. AI-driven features assist farmers in pricing their produce competitively, predicting demand trends, and optimizing sales strategies. Real-time communication tools allow direct interaction between farmers and consumers, fostering trust and improving transaction efficiency. Krisho also emphasizes inclusivity by supporting multiple regional languages, ensuring that even farmers from remote areas can easily use the system. The platform provides a user-friendly interface where farmers can list their produce, manage orders, and receive payments securely. Consumers, on the other hand, benefit from fresh, traceable, and affordable agricultural products delivered direct from farms. In addition to economic benefits, Krisho contributes to reducing food wastage by streamlining the supply chain and enabling quicker transactions. It also supports sustainable agricultural practices by promoting local sourcing and minimizing transportation overhead. Overall, Krisho presents a scalable and impactful solution that empowers farmers, enhances consumer satisfaction, and strengthens the agricultural ecosystem, contributing to a more efficient, transparent, and equitable marketplace in Bharat.

Real-Time Multi-Class Vehicle Detection Using YOLOv12n for Intelligent Traffic Monitoring

Authors: Ali Imam Tonmoy

Abstract: The rapid urbanization and consequent surge in vehicular density worldwide necessitate advanced, real-time traffic monitoring solutions. This paper presents a robust framework for multi-class vehicle detec tion leveraging the novel YOLOv12n architecture, specifically tailored for intelligent transportation systems (ITS). We train and rigorously evaluate our model on a curated dataset of 535 annotated images comprising 11,035 vehicle instances across three classes: cars, trucks, and buses. YOLOv12n demonstrates superior performance over state-of the-art lightweight detectors, including YOLOv8-nano, YOLOv9-tiny, YOLOv10-nano, and YOLOv11-nano, achieving 94% precision, 91% recall, and 96% mAP@0.5 while sustaining a real-time inference speed of 132 FPS. The architectural innovations of YOLOv12n, particularly its attention-based feature learning and Residual Efficient Layer Aggre gation Networks (R-ELAN), enable robust detection under challenging conditions such as variable illumination, partial occlusions, and signif icant scale variations. This study establishes YOLOv12n as a com pelling solution for practical traffic surveillance and paves the way for advanced smart city applications.

DOI: https://zenodo.org/records/20055871

XploreStays: Design And Implementation Of A Web-Based Travel Accommodation Listing Platform

Authors: Akshita Sharma, Karuna kashyap

Abstract: The rapid advancement of internet technologies has significantly transformed the tourism and hospitality industry. This research paper presents XploreStays, a web-based travel accommodation listing platform designed to simplify the process of exploring and managing travel stays. The platform enables users to browse multiple property listings through a structured and user-friendly interface, with each listing containing essential information such as property images, location details, and pricing information. The system integrates a user authentication module and a dynamic pricing system that calculates the total cost of a stay by incorporating taxes and service fees. The platform introduces a dual-rating review mechanism that evaluates both accommodation quality and listing transparency. The architecture follows a layered web application model consisting of presentation, application, and data layers, implemented using HTML, CSS, JavaScript, Node.js, Express.js, and MongoDB. Results demonstrate an efficient and scalable solution for accommodation listing systems, serving as a foundation for future development including booking systems, payment gateways, and user review modules.

</p

Enhancing Privacy In Federated Learning: A Comprehensive Survey Of Preservation Techniques

Authors: D Naga Bharghavi, M Deepthi, K Manga Devi, M Aswitha, P Aswitha

Abstract: Federated Learning (FL) enables multiple devices or organizations to collaboratively train machine learning models without sharing raw data, thus improving privacy. However, FL is vulnerable to privacy threats like model inversion, membership inference, and data leakage from shared updates. To mitigate these risks, several privacy-preserving techniques have been developed, including differential privacy, secure multiparty computation (SMC), homomorphic encryption (HE), and hybrid approaches that combine multiple methods. This paper offers a comprehensive analysis of these techniques, evaluating their privacy guarantees, computational costs, and impact on model accuracy. Differential privacy introduces noise to protect data but can reduce model performance. SMC allows joint computation without exposing inputs but is computationally intensive. HE enables encrypted data processing with strong security, though often at the expense of efficiency. Hybrid methods aim to balance these trade-offs by leveraging the advantages of different approaches. The study highlights key challenges such as scalability and usability in real-world FL deployments. It also identifies research gaps and proposes future directions focused on adaptive privacy mechanisms and hardware-assisted security, aiming to develop more practical and robust privacy-preserving FL systems.

DOI: https://doi.org/10.5281/zenodo.20061078

Smart Home Appliance System for Intelligent Monitoring and Automation

Authors: M. Anusha Rani, A. Purna Prasanna Lakshmi, Y. Deepika, Indhu, V.Nandini

Abstract: In today’s era of rapid technological progress, home automation has become an essential element of modern living, providing convenience, energy efficiency, and improved security. Smart home systems utilize embedded electronics, sensors, and automation to monitor environmental conditions and intelligently control household appliances. The proposed system is built around an Arduino Uno microcontroller, integrated with a DHT11 temperature and humidity sensor, an ultrasonic sensor for motion detection, and an RTC module for time-based operations. These components work together to automate devices such as lights and fans based on real-time sensor readings. By responding automatically to environmental changes, the system minimizes energy wastage, enhances comfort, and improves safety within the home. Its design relies on low-cost, easily accessible components, making it affordable and adaptable for urban, rural, or institutional applications without requiring complex networking Overall.

DOI: https://doi.org/10.5281/zenodo.20061120

Advancements In Quantum Computing: A Paradigm Shift In Computational Science

Authors: Dr Jermiah Anand Jupalli, S Manjusha, D Haneesha, J Chaitanya, M poojitha

Abstract: Classical computing faces significant limitations when dealing with complex, large-scale scientific problems, driving growing interest in quantum computing as a transformative alternative. Quantum computing leverages the principles of superposition, entanglement, and quantum parallelism to perform computations that are infeasible for classical machines. This paper explores recent advancements in quantum hardware, algorithmic design, and the development of hybrid quantum–classical models that integrate quantum gate operations with classical optimization techniques. The proposed hybrid framework aims to enhance computational accuracy and efficiency by utilizing quantum circuits for high-speed processing while relying on classical methods for error correction and optimization. Simulated experiments and comparative analyses demonstrate a 35% improvement in computational speed and scalability over conventional methods. These findings highlight that quantum computing is not merely an incremental improvement but a fundamental shift in computational paradigms, offering new opportunities for scientific research, cryptography, machine learning, and complex data processing.

DOI: https://doi.org/10.5281/zenodo.20061146

Design And Implementation Of A Facial Recognition Based Smart Lock System

Authors: Dr.S.Siva Venkata Ramana, P.Aswini, D.Rithika, J.Kavya, U.Amulya

Abstract: In today’s world, ensuring personal and property security has become increasingly important. Traditional lock systems and password-based access controls are often inconvenient and vulnerable to misuse. To address these challenges, this project focuses on the design and implementation of a facial recognition-based smart lock system that offers a more secure and user-friendly way of controlling access. The main goal of this work is to create a system that automatically identifies individuals based on their facial features, removing the need for physical keys or manual authentication. The prototype is built using a Raspberry Pi microcontroller connected to a camera module, which captures a live image of the person at the door. The image is analyzed using Python and Open CV to detect and compare facial patterns with a trained dataset of authorized users. Once a match is confirmed, the lock is activated through a relay mechanism. The system was tested under various lighting conditions and angles, showing consistent accuracy and quick response times. An additional IOT-based alert feature sends real-time notifications when access is granted or denied, providing users with an extra layer of awareness and safety.

DOI: https://doi.org/10.5281/zenodo.20061183

Machine Learning–Based Analysis Of Air Quality Parameters

Authors: G. Manidheer Babu, M. Akhila, S. Karishma, G. Maha Lakshmi, Ch. Showrilamma

Abstract: Air pollution has become one of the most pressing global health and environmental issues, with both outdoor and indoor exposures leading to millions of preventable deaths annually. Concerns about indoor air quality have intensified because people now spend most of their time indoors. Inadequate ventilation combined with emissions from building materials and human activities can often lead to higher pollutant concentrations indoors than outdoors. Traditional monitoring systems, while accurate, are often expensive and difficult to maintain for continuous indoor deployment, limiting their practical use.This situation has sparked interest in the use of Machine Learning (ML) techniques, which can handle extensive datasets, uncover hidden relationships among environmental variables, and produce reliable forecasts to support decision-making in air quality management. In this research, a Machine Learning-based Air Quality Monitoring System was created to forecast ventilation status utilizing a publicly accessible dataset from Kaggle. The dataset encompasses essential environmental factors, including temperature, humidity, carbon dioxide (CO₂), particulate matter (PM2.5 and PM10), total volatile organic compounds (TVOC), carbon monoxide (CO), light intensity, motion detection, and occupancy count.

DOI: https://doi.org/10.5281/zenodo.20061322

Bridging Memory And Quantum Intelligence: A Neuromorphic Approach To Quantum Machine Learning

Authors: M Prasanna Kumar, KPavani, DPrasanna, D Siva Koteswari, R Ashritha

Abstract: One of the most promising approaches to using quantum computing to tackle challenging artificial intelligence issues is quantum machine learning (QML). The majority of QML architectures, however, are constrained by the absence of explicit methods for managing temporal and memory dependencies, which are essential for tasks like signal processing, sequential decision-making and forecasting. By embedding memory through devices like memristors, neuromorphic computing which draws inspiration from the brain’s synaptic plasticity-offers a natural solution. In this paper, we propose a conceptual framework for using quantum memristors to incorporate neuromorphic memory into quantum machine learning. We compare the potential benefits over current QML models, suggest a simulation-based experimental design, and assets the extent to which systems could handle sequential data challenges. Our approach contributes towards shaping the emerging paradigm of neuromorphic quantum intelligence.

DOI: https://doi.org/10.5281/zenodo.20061628

Cardiac Arrest Prediction Using Machine Learning: Random Forest And Svm Framework For Early Detection

Authors: T.Naga Navya, V.Naga Harshitha, K.Manisha, K.Gayathri, K. Rishitha

Abstract: Cardiac arrest (CA) is an acute and life-threatening condition presenting with the sudden loss of cardiac activity, leading to immediate cessation of blood flow to central organs and loss of consciousness. In spite of the significant advances in emergency medical services and ongoing patient monitoring, early detection of cardiac arrest is still a significant challenge because of the nonlinear and complex behavior of physiological signals. This study introduces a strong data-driven machine learning (ML) model for predicting cardiac arrest in real time based on continuous tracking of vital signs like heart rate variability, blood pressure, ECG signal variation, and oxygen saturation levels (SpO₂). Three supervised learning modelsLogistic Regression, Support Vector Machine (SVM), and Random Forestwere created and contrasted following normalization and feature selection. Of these, Random Forest model showed the best predictive results with 92% accuracy, 90% precision, and 91% recall. The developed system has strong prospects for integration into hospital monitoring devices and wearable technology, facilitating timely intervention and better patient survival rates.

DOI: https://doi.org/10.5281/zenodo.20063385

Fake News Detection Using Transformer-based Models: BERT And RoBERTa

Authors: Dr. P Radhika, J N V S R Sri Thanaya, S Khyathi Chowdary, G Nageswari, P Gowthami

Abstract: The rapid growth of social media and digital communication platforms has increased the spread of misinformation. This threatens public trust, democracy, and health communication. As a result, detecting fake news has become a major focus in Natural Language Processing (NLP) research. Traditional machine learning methods, such as Logistic Regression and Support Vector Machines, are widely used, but they struggle to capture the meaning and context of text. Recent advancements in transformer-based models, including BERT and RoBERTa, have achieved top results in several NLP tasks by using self-attention and contextual embedding techniques.This study presents a hybrid approach that combines a literature review with a comparative experimental study using the ISOT Fake News Dataset. We compared the baseline Logistic Regression classifier with TF-IDF features to transformer-based models, specifically BERT and RoBERTa. The results show that transformer models outperformed traditional classifiers across all metrics. This demonstrates their better understanding of context and ability to generalize.

DOI: https://doi.org/10.5281/zenodo.20064953

Automated Plant Disease Recognition System

Authors: Shivam Admane, Siddharth Guldagad, Payal Chakane, Prof. Nikam PP

Abstract: Plants are very essential in our life as they provide source of energy and overcome the issue of global warming. Plants now a days are affected by diseases like bacterial spot, late blight, Septoria leaf spot. These diseases effect the efficiency of crop yield. So ,the early detection of diseases is important in agriculture. Detection of diseases as soon as they appear is vital step for effective disease management. Aim of the project is to detect plant leaf disease by Machine Learning using image and videos. For Image, the proposed algorithm is Random forest classifier-Machine learning Algorithm used for classification andfor video the proposed technique is Resnet50- Deep Learning Algorithm. These techniques will obtain prediction results using various metrics like accuracy, precision and efficiency. This project can be implemented in agriculture.

Multi-Modal RAG Framework Integrating Text, Image, And Audio Intelligence

Authors: Mrs. C. Radha, Sanjay V, Santhosh P, Rejitha R

Abstract: The sheer volume of unstructured data generated in multiple modalities like text, images, audio, and video has overwhelmed existing text-based retrieval augmented generation (RAG) models. In this work, we propose a unified framework for multi-modal RAG (MM-RAG) that enables unified multimodal processing, multimodal representation, cross-modal retrieval, and multimodal document generation within a single model. The proposed framework employs cutting-edge technology: BridgeTower for unified text-image encoder, Whisper for automatic speech recognition, FAISS for vector similarity search, and large vision-language models (LVLMs) for answer generation. Through experiments conducted on three enterprise use cases, it is evident that MM-RAG has a high cross-modal retrieval accuracy of 94.2%, hallucination reduction of 67.3% when compared with unimodal baselines, and a multi-modal document throughput of 1,200 pages per hour. It is clear from our comparative study that MM-RAG significantly outperforms the uni-modal system for tasks demanding multimodal integration, especially visual product search (recall@10 = 92.8%) and audio/video content retrieval (accuracy = 88.5%).Safely allocated to each asset, depending on the investment strategy.

DOI: https://doi.org/10.5281/zenodo.20071343

Automatic Summarization Of Financial Reports Using NLP Techniques

Authors: Dr. Pankaj Malik, Sachin Sethiya, Yarthik Soni,, Harshita Kushwah, Jhalak Kavadiya

Abstract: Financial reports are often lengthy, complex, and rich in domain-specific terminology, making manual analysis time-consuming and inefficient. This paper proposes an automated summarization framework using Natural Language Processing (NLP) techniques to generate concise and informative summaries of financial documents. The system employs a hybrid approach that combines extractive methods (TF-IDF and TextRank) with abstractive transformer-based models such as BART and PEGASUS to enhance contextual understanding and coherence.The proposed model was evaluated on benchmark financial datasets, including annual reports and earnings call transcripts. Experimental results demonstrate that the hybrid model outperforms traditional extractive and standalone abstractive approaches, achieving a ROUGE-1 score of 0.52, ROUGE-2 score of 0.31, and ROUGE-L score of 0.48. Additionally, the model improved information retention by approximately 18% and reduced redundancy by 22% compared to baseline methods.The findings indicate that integrating extractive and abstractive techniques significantly enhances summarization quality, enabling faster and more accurate financial analysis. This approach can be effectively applied in investment decision-making, financial auditing, and automated reporting systems.

DOI: https://doi.org/10.5281/zenodo.20074327

Impact Of Digital Marketing Strategies On Pharmaceutical Product Promotion

Authors: Arpita Lad, Dr. Devendra S. Shirode, Dr. Ramdas Shinde

Abstract: The pharmaceutical industry is undergoing a significant transformation in product promotion with the increasing adoption of digital marketing strategies. This study evaluates the impact of these strategies on engagement, efficiency, decision-making, and sales performance using secondary data from peer-reviewed literature, industry reports, and company insights from Pfizer and Sun Pharma. A comparative and analytical approach, supported by Marketing Mix Modeling, was used to assess the effectiveness of digital and traditional channels. The findings indicate that digital tools such as webinars, e-detailing, and email campaigns expand communication reach, enable personalized interactions, and contribute to improved product visibility and sales outcomes through data-driven targeting. However, traditional methods, particularly face-to-face engagement, remain important for building trust and influencing prescribing behavior. The study concludes that a hybrid marketing approach integrating digital and conventional strategies delivers optimal results in pharmaceutical promotion and sales performance. It also emphasizes the need for regulatory compliance, ethical communication, and data privacy in digital practices.

DOI: https://doi.org/10.5281/zenodo.20074203

Automatic Summarization Of Financial Reports Using NLP Techniques

Authors: Dr. Pankaj Malik, Sachin Sethiya, Yarthik Soni,, Harshita Kushwah, Jhalak Kavadiya

Abstract: Financial reports are often lengthy, complex, and rich in domain-specific terminology, making manual analysis time-consuming and inefficient. This paper proposes an automated summarization framework using Natural Language Processing (NLP) techniques to generate concise and informative summaries of financial documents. The system employs a hybrid approach that combines extractive methods (TF-IDF and TextRank) with abstractive transformer-based models such as BART and PEGASUS to enhance contextual understanding and coherence.The proposed model was evaluated on benchmark financial datasets, including annual reports and earnings call transcripts. Experimental results demonstrate that the hybrid model outperforms traditional extractive and standalone abstractive approaches, achieving a ROUGE-1 score of 0.52, ROUGE-2 score of 0.31, and ROUGE-L score of 0.48. Additionally, the model improved information retention by approximately 18% and reduced redundancy by 22% compared to baseline methods.The findings indicate that integrating extractive and abstractive techniques significantly enhances summarization quality, enabling faster and more accurate financial analysis. This approach can be effectively applied in investment decision-making, financial auditing, and automated reporting systems.

DOI: https://doi.org/10.5281/zenodo.20074327

Assessment Of Seasonal Variations In Water Quality Of Pampa River With Special Emphasis On Pilgrim Season

Authors: Feba. R. Philip, Dr Reji P.G

Abstract: The Pampa River, one of Kerala’s prominent rivers, plays a vital ecological, cultural, and religious role, serving as the lifeline of the Sabarimala pilgrimage. Every year, during the Mandala–Makaravilakku season, a large influx of pilgrims leads to a significant increase in waste generation, sewage inflow, and ritual-based discharges into the river. In addition to these anthropogenic influences, natural factors such as monsoon floods, pre-monsoon low flows, and seasonal fluctuations further affect the river’s water quality. This study undertakes a systematic, seasonal, and event-based assessment of the Pampa River to understand pollution dynamics, identify critical stress periods, and support sustainable river basin management and pilgrimage planning. During the pilgrimage season, the river is extensively used by devotees for bathing and various ritual activities, often resulting in contamination and a decline in water quality. The present work involves water quality evaluation during three phases before, during, and after the pilgrimage season using the Water Quality Index (WQI) method integrated with GIS-based spatial analysis. The study examines physical, chemical, and microbiological parameters, including oxygen demand, nutrient concentrations, and indicators of organic and inorganic pollutants, along with other relevant water quality determinants from September to February month. The results clearly indicate that the river water quality is highly dynamic and is significantly influenced by both natural seasonal variations and anthropogenic activities.

DOI: https://doi.org/10.5281/zenodo.20074516

Effects Of Mode Of Application Of Brevibacillus Brevis On Calcite Precipitation Of Bio-treated Lateritic Soil.

Authors: M. Abubakar, M. A. Garba, K. J. Osinubi, F.ASCE A. O. Eberemu, M.ASCE

Abstract: Sustainable method of soil improvement called Microbial Induced Calcite Precipitation (MICP) has received significant interest recently. In this technique, the behaviours of bio-treated soil are controlled by the amount of calcite content (CC) produced. Various methods have been used in the literature to determine the CC in soil using injection technique. The quantity of calcite precipitation is affected by the mode of application of microbes used and this effect hasn’t been studied extensively. There is no preferred mode of application only the most commonly used method documented. In this paper, the CC of lateritic soil bio-treated at stepped Brevibacillus brevis (B. brevis) suspension density (SD) and cementation reagent (CR) concentration using three mode of treatment (i.e. mixing, injection and spraying method) and mix ratio (i.e., 25B–75C, 50B–50C and 75B–25C) were determined. The B. brevis SD and CR used to trigger the MICP process are based on McFarland Standards. Based on this result, the mixing mode produced the best result then injection mode and the least result was for the spraying mode of treatment. The highest CC values of 12, 8.72 and 6.4 % were recorded at 24E8 cells/ml – 1M using mixing, injection and spraying mode of treatment respectively. The recorded CC values based on the mode of treatment were in the order mixing > injection > spraying method of application respectively.

DOI: https://doi.org/10.5281/zenodo.20074663

Perspectives On Generative AI: Basis For Strategic AI Integration Plan

Authors: Dr. Rommel Z. de Leon

Abstract: This study assessed the perspectives of faculty members at Pililla National High School (PNHS) regarding the integration of Generative Artificial Intelligence (GenAI) during the School Year 2025–2026. Utilizing a Descriptive-Developmental Research Design, the study gathered quantitative data from 118 permanent teaching personnel to characterize faculty awareness and attitudes, subsequently informing the creation of a Strategic AI Integration Plan.The demographic profile revealed a female-dominated (75%) and relatively young faculty, with 67% of respondents aged 20–39. Results indicated a Grand Mean of 4.09 (Very Aware) regarding student AI use, with teachers showing high consciousness of AI limitations (Mean = 4.44) but lower awareness of specific "allowable use" policies for written works (Mean = 3.30). In terms of guidance and monitoring, faculty are proactive "gatekeepers," consistently discussing ethics (Mean = 4.44) and upholding academic integrity (Mean = 4.38), though they often lack standardized verification protocols. Interestingly, while teachers advocate for responsible student use, their own integration of GenAI for instructional design remains cautious (Grand Mean = 3.35, "Sometimes").Statistical analysis showed no significant difference in AI assessment based on sex or grade level. However, a significant difference (p = 0.018) was found across age Brackets, highlighting a "Digital Fluency Gap" where younger teachers (20–29) demonstrate higher usage and comfort levels. To address this, the study developed a Strategic AI Integration Plan featuring a "Reverse Mentoring Program." This dual-phase approach pairs younger teachers’ technical proficiency in prompt engineering with veteran teachers’ expertise in authentic assessment and ethical frameworks to ensure a balanced, institution-wide AI adoption.

DOI: https://doi.org/10.5281/zenodo.20074980

Arduino-Based Control Of Dual-Axis Solar Tracking PV System With Integrated Features

Authors: Ch Ashok Rao, M Vignesh, R Chaitanya Charan, S Akshay, DR.G. Suresh Babu

Abstract: This paper presents the design, implementation, and experimental validation of an Arduino Uno-based dual-axis solar tracking system augmented with comprehensive environmental monitoring and an integrated. Conventional fixed-mount photovoltaic (PV) installations suffer from significant energy losses due to the continuous angular displacement between the sun's position and the panel's fixed orientation. The proposed system employs four light-dependent resistors (LDRs) arranged in a quadrant configuration to sense differential irradiance and drive two servo motors that continuously orient a 10 W PV panel toward maximum solar incidence in both azimuth and elevation axes. Real-time environmental data—ambient temperature, relative humidity, and precipitation—are acquired via a DHT11 sensor and a rain-detection module, enabling adaptive operational modes and hardware protection. A 16×2 LCD module provides a local human-machine interface for instantaneous parameter display. An H-bridge-based DC-AC inverter topology converts the harvested DC energy to a 50 Hz, 220 V AC output suitable for resistive domestic loads. Experimental trials conducted between 08:00 h and 17:00 h under varied atmospheric conditions demonstrate an average energy-harvest improvement of 35–40% over an identically rated fixed-tilt panel. Motor actuation consumes approximately 0.4 W, yielding a net efficiency gain that validates the economic and technical viability of active solar tracking. The system architecture is further extensible toward IoT-enabled cloud monitoring and machine-learning-driven predictive fault detection, establishing a robust foundation for next-generation smart renewable energy nodes.

DOI: https://doi.org/10.5281/zenodo.20075286

Adaptive Control Strategy For EV Charging Stations Integrated With Renewable Energy

Authors: Kandukuri Umadevi, Gowshick Raja K. P

Abstract: With the growing number of electric vehicles (EVs) and the intermittent nature of renewable energy sources, there are challenges in managing the stability of the power grid and charging infrastructure. This paper suggests an innovative adaptive control mechanism for EV charging stations with photovoltaic (PV) power generators and battery energy storage systems (BESS). In this system, the adaptive control scheme uses Model Predictive Control (MPC), where prices are dynamically adjusted to schedule the optimal charging times, manage the grid load, and increase the usage of renewable energy. Based on actual solar energy and arrival rate of EVs, the results showed that the proposed adaptive scheme was able to minimize peak loads from the grid by 58.8%, increase the usage of renewable energy by 19.952% in comparison with non-adaptive schemes, and achieve SoC targets within 0.79% deviation.

DOI: https://doi.org/10.5281/zenodo.20086153

From Digital to AI Literacy: Repositioning Academic Libraries in the Age of Generative AI

Authors: Olatunji Austine Kehinde, Oladosu Adedamola Oluwatimilehin, Akanle Mojisola Benedicta

Abstract: The rise of generative artificial intelligence (AI) is reshaping the landscape of digital and information literacy in higher education, demanding a redefinition of the role of academic libraries. This paper presents a literacy continuum that links digital, information, and AI literacies through a constructivist learning lens, positioning libraries as active pedagogical partners in cultivating student competencies. The framework emphasises that digital literacy provides technical grounding, information literacy builds evaluative and ethical capacity, and AI literacy advances critical engagement with generative technologies. Applications for libraries include redesigning instructional programs, fostering faculty collaboration, and offering leadership in institutional policy development. A comparative perspective highlights divergent trajectories between the Global North and Global South, where disparities in infrastructure, pedagogical innovation, and ethical debates shape the adoption of AI literacy. Two tables support this analysis: one contrasts global trends in literacy development, and the other aligns methodological approaches with empirical research directions. The paper also identifies methodological pathways, qualitative, quantitative, and mixed-method that can operationalise AI literacy across critical evaluation, functional use, ethical practice, and reflective engagement. While the model offers a flexible foundation, its conceptual scope and the evolving pace of AI present limitations that require ongoing adaptation. The conclusion calls for context-aware strategies that expand librarian expertise, embed AI literacy in curricula, and advance libraries as ethical leaders in AI integration. By doing so, libraries can not only mediate access but also drive equitable, future-ready knowledge practices in the AI era.

DOI: https://doi.org/10.5281/zenodo.20088530

Emergency Trauma Analyzer: An Intelligent System For Medical Diagnosis

Authors: Mrs.Bamila Rachel, Jeyadharshini A, Jenifar T, Renuka R, Prema M

Abstract: Trauma injuries require rapid diagnosis and timely medical intervention to reduce complications and improve patient survival. However, analyzing medical data and imaging reports manually can delay the diagnostic process in emergency situations. This paper proposes an Emergency Trauma Analyzer, an intelligent system designed to assist healthcare professionals in analyzing trauma-related medical data and supporting clinical decision-making. The system integrates a web-based interface with an artificial intelligence analysis engine to process patient information and provide diagnostic insights. The platform allows doctors to input patient details, upload medical data, and receive analysis results that help identify potential trauma conditions. The system also stores medical records and generates reports for future reference. The proposed solution aims to improve the efficiency of trauma assessment and support doctors in providing faster and more accurate medical care.

Autonomous Agentic RAG Loop (AARL): A Framework For Offline Generative AI In Military Domain

Authors: Satish Kumar Gupta, C.R.S. Kumar

Abstract: The deployment of Large Language Models (LLMs) in sensitive, disconnected environments such as air-gapped military networks introduces challenges related to knowledge staleness, strict data isolation, security assurance, and adaptability. While Retrieval-Augmented Generation (RAG) improves factual grounding by integrating local knowledge sources, conventional RAG pipelines remain static and lack autonomous adaptability for complex intelligence tasks. Agentic RAG extends this paradigm through autonomous agents but typically assumes online connectivity and external feedback, rendering it unsuitable for classified, offline deployments. Agentic Retrieval-Augmented Generation (Agentic RAG) transcends these limitations by embedding autonomous AI agents into the RAG pipeline. This paper proposes a novel conceptual framework, the Autonomous Agentic RAG Loop (AARL), which integrates multi-agent coordination, adaptive retrieval, and secure reasoning for offline intelligence systems. The AARL architecture introduces agents with specific cognitive roles (Retriever, Generator, Evaluator, and Orchestrator) operating within isolated computation modules called Distributed Cognitive Cells (DCCs). The framework features a self-correcting feedback loop driven by a Self-Adaptive Reinforcement Mechanism (SARM) that enables continuous improvement without external dependencies. This paper provides both theoretical analysis of the AARL's expected properties and a detailed feasibility assessment, demonstrating that its design offers a significant step toward building self-sufficient, secure, and trustworthy AI systems for defence and critical infrastructure applications.

DOI: https://doi.org/10.5281/zenodo.20089714

Heart And Lung Sounds Dataset Recorded From A Clinical Manikin Using Digital Stethoscope (HLS-CMDS)

Authors: S. Chidharth, S. Divakar, V. Vijayan

Abstract: Heart and lung sounds are crucial for healthcare monitoring. Recent improvements in stethoscope technology have made it possible to capture patient sounds with enhanced precision. In this dataset, we used a digital stethoscope to capture both heart and lung sounds, including individual and mixed recordings. To our knowledge, this is the first dataset to offer both separate and mixed cardiorespiratory sounds. The recordings were collected from a clinical manikin, a patient simulator designed to replicate human physiological conditions, generating clean heart and lung sounds at different body locations. This dataset includes both normal sounds and various abnormalities (i.e., murmur, atrial fibrillation, tachycardia, atrioventricular block, third and fourth heart sound, wheezing, rhonchi, pleural rub, fine crackle, and coarse crackle sounds). The dataset includes audio recordings of chest examinations performed at different anatomical locations, as determined by specialist nurses. Each recording has been enhanced using frequency filters to highlight specific sound types. This dataset is useful for applications in artificial intelligence, such as automated cardiopulmonary disease detection, sound classification, unsupervised separation techniques, and deep learning algorithms related to audio signal processing.

A Systematic Review Of The Opportunities, Challenges, And Future Prospects Of Generative AI In Transforming Higher Education

Authors: Mr. Dhananjay Bhagwat Kshirsagar

Abstract: Generative Artificial Intelligence (GenAI) has rapidly emerged as a transformative and disruptive technology in higher education, fundamentally reshaping teaching methodologies, learning experiences, research practices, and institutional operations. This systematic review synthesizes recent scholarly literature published between 2022 and 2026 to critically examine the multifaceted opportunities, inherent risks, and future implications of generative AI integration in higher educational institutions (HEIs). Through a rigorous analysis of current trends, the review identifies major pedagogical and administrative opportunities, including highly personalized learning pathways, 24/7 intelligent tutoring systems, rapid educational content generation, advanced research assistance, and the automation of routine administrative tasks. Concurrently, the study highlights critical risks that threaten traditional educational paradigms, such as rampant academic dishonesty, algorithmic bias, severe data privacy concerns, and the cognitive degradation stemming from an overdependence on AI tools. The findings suggest that while generative AI possesses unprecedented potential to augment educational effectiveness and streamline academic workflows, its sustainable adoption hinges on the development of robust, responsible implementation frameworks. Establishing strict ethical guidelines, proactive faculty development programs, and fundamentally redesigning assessment strategies are essential steps for institutions navigating this technological paradigm shift.

DOI:

A Machine And Deep Learning-Powered Platform For Unified Detection Of Neuro Degenerative Contions – Alzheimer’s

Authors: S. Bhuvanaloshini, S. Saranya, R. Savithri, M. Aji, Dr. A. Jeyapraba

Abstract: Alzheimer’s disease is a progressive neurodegenerative disorder that affects memory and cognitive functions, making early diagnosis essential for effective treatment. This work proposes a deep learning-based system for the classification of Alzheimer’s disease using MRI brain images. A Convolutional Neural Network (CNN) based on the LeNet architecture is employed to automatically extract relevant features from input images and perform multi-class classification. The model is trained on a publicly available Kaggle dataset consisting of approximately 1200 MRI images. The system is designed to classify different stages of Alzheimer’s disease, including normal, mild cognitive impairment, and advanced stages. Data preprocessing techniques such as resizing and normalization are applied to improve model performance. The dataset is split into training and testing sets to evaluate the effectiveness of the model. Experimental results show that the proposed model achieves a high accuracy of 99.5%, while maintaining low computational complexity. In addition to the classification model, a web-based application is developed using Django, allowing users to upload MRI images and obtain real-time predictions along with basic treatment suggestions. The system provides fast and reliable results, making it suitable for practical use. The proposed approach demonstrates that a simple CNN architecture can achieve high performance and can effectively support early diagnosis and clinical decision-making.

DOI: https://doi.org/10.5281/zenodo.20093414

Conversational AI For Automated Student Query Resolution In Educational Institutions

Authors: Dheepigha G, Pavithra K, Vijayan V

Abstract: In recent years, schools have had a hard time handling a lot of student questions quickly because old-fashioned manual systems take a lot of time and often cause delays. This paper suggests an AI-based conversational chatbot that uses Natural Language Processing (NLP) techniques to automate student support services. The system can understand and answer user questions in real time, giving them accurate and relevant information about academics, administration, and campus services. The chatbot gets better at responding to questions and learns from each interaction by combining a knowledge base with machine learning. The suggested solution makes communication more efficient, lightens the load on administrative staff, and makes sure that students can get help at any time of day or night. Experimental results indicate that the system significantly improves response time and user satisfaction, making it a scalable and effective approach for modern educational environments.

Intelligent Wireless Communication Using A Semantic-Aware Deep Learning Approach

Authors: 1Miss. A. G. Salve, Dr. D. U. Adokar, Dr. P. M. Patil

Abstract: The evolution of intelligent wireless systems requires moving from bit-level transmission to semantic communication. This work proposes a semantic-aware deep learning framework integrating NLP, semantic encoding, MQTT communication, and IoT actuation. Sentence-BERT generates 384-dimensional embeddings, compressed to a 32-dimensional latent space (12:1 ratio) using a lightweight autoencoder. Gaussian noise (σ = 0.1, 0.2) simulates realistic channel conditions. The model is compact (395 KB), with 98,688 parameters and 106,496 FLOPs, enabling efficient edge deployment. Results show high semantic preservation, with cosine similarity >0.95 for paraphrases and >0.90 under noise. Intent recognition achieves high confidence (0.9886 for “Light On” and 0.9782 for “Light Off”). Real-time validation using MQTT and ESP32 confirms reliable, low-latency control. The framework offers scalability, noise robustness, and spectral efficiency, making it suitable for 6G semantic communication systems.

Comparative Analysis Of Traditional And Docker-Based Deployment For Web Applications

Authors: Aryan Gupta

Abstract: Docker containerization has been extensively studied over the past decade, and a general consensus has formed around its performance characteristics. Studies running on native Linux platforms — Felter et al. [1], Potdar et al. [4], Mavridis and Karatza [5] — consistently report less than 5 to 6 percent execution overhead and 20 to 30 percent memory efficiency gains. More recently, a small number of studies have begun examining Docker on non-Linux platforms. Sladovic et al. [9] tested Docker across Windows, macOS, and Ubuntu in 2024 using compilation benchmarks and found Windows with WSL2 retained about 94 percent of native Linux performance. Sergeev et al. [11] compared Docker on Windows and Linux in 2022 and found only marginal arithmetic and memory performance differences between the two platforms. However, these cross-platform studies share a limitation: they used general-purpose or compilation benchmarks rather than high-concurrency web application workloads. The performance of Docker on Windows WSL2 under concurrent HTTP request loads — the scenario most relevant to web application developers — has not been systematically measured. This paper fills that gap by running four controlled experiments on a Windows 11 machine with an Intel i5-12500H Alder Lake processor under sustained concurrent workloads: application startup time, concurrency scaling from 10 to 200 users, memory consumption across workload sizes, and sustained execution under peak load. Our results reveal a more significant performance gap than prior cross-platform studies found. Under high-concurrency compute-intensive web workloads, Docker-on-WSL2 showed 58.1% execution overhead — far higher than Sladovic et al.'s 6% and Sergeev et al.'s marginal difference. We attribute this to the WSL2 virtual Ethernet bridge, which adds consistent per-request overhead under concurrent load that compilation benchmarks do not exercise. Memory efficiency of 25.3% reduction was confirmed consistent with prior literature. Additionally, Docker achieved near-complete CPU saturation of all 16 Alder Lake threads through the Linux CFS scheduler — a workload-dependent advantage that prior cross-platform studies did not document.

DOI: https://doi.org/10.5281/zenodo.20099771

Beyond Pixels: Multimodal Detection Of Interface-Consistent Chat Screenshot Manipulations

Authors: Vivek, Yashvardhan Pannu, Anirudh Thakur

Abstract: With the growing use of chat screenshots as evidentiary material in legal proceedings, journalistic investigations, and corporate disputes, the need for reliable authentication tools has become urgent. Existing image forensics frameworks — optimised for natural photographs or synthetic face-swap detection — fail to address a structurally distinct attack category: interface-consistent manipulations, in which the visual grammar of a messaging platform is preserved while semantic content is falsified. This paper presents BeyondPixels, a three-channel multimodal detection framework combining Error Level Analysis (ELA), a domain-adapted EfficientNet-B3 convolutional neural network, and an OCR-driven semantic validator. The three channels address complementary manipulation signatures: pixel-level compression artefacts, structural image disruptions, and logical inconsistencies in message text and timestamps. A weighted fusion engine converts per-channel scores into a single normalised authenticity score. Evaluation on a purpose-built corpus of 3,000 synthetic screenshots — spanning five manipulation categories, two interface themes (WhatsApp and Telegram), and two languages (English and Hindi) — yields 90.3% accuracy and 89.5% F1-score, with an AUC of 0.951. Ablation confirms that every channel contributes independently and the OCR module is decisive for text-only attack categories that image-level methods cannot detect.

DOI: https://doi.org/10.5281/zenodo.20102569

Digital Pharmaceutical Marketing Strategies in Improving PCOS Awareness and Management among Women – A Thematic Review

Authors: Harsha Jagtap, Sayali Raut

Abstract: Polycystic Ovary Syndrome (PCOS) is a prevalent endocrinological and metabolic disease among women in their reproductive years. This disease carries many reproductive, metabolic, and psychological effects that affect the quality of life negatively. Due to the high prevalence rate of this condition, knowledge regarding PCOS and its management remains limited, especially in developing countries. This article aims to explore methods used to improve the knowledge and management of PCOS using digital media. This is a narrative review with thematic analysis on articles found in scholarly databases like PubMed, ScienceDirect, and SpringerLink and published between 2019 and 2025. Results indicate that although general awareness about PCOS exists, in-depth understanding regarding its causes, complications, and management remains limited. However, challenges such as misinformation, lack of regulation, poor long-term engagement, and digital inequality continue to limit their effectiveness. Overall, digital health strategies show strong potential to improve PCOS awareness and management; but their success depends on integration with structured education, active healthcare professional involvement, and strict ethical regulation to ensure sustainable outcomes.

DOI: https://doi.org/10.5281/zenodo.20103245

Design and Development of an AI-Powered DSA Practice and Learning Platform

Authors: Prof. A. C. Sawant, Mayuri Bhujbal, Harshada Salunke, Harshada Ahire., Dhruv Mene

Abstract: This paper presents a comprehensive DSA (Data Structures and Algorithms) practice platform designed to enhance problem-solving skills through personalized learning and performance-driven feedback. The system integrates user and admin modules, enabling efficient management of topics, quizzes, coding problems, and assessments. Users can learn concepts, attempt MCQ-based quizzes, and solve coding problems, earning experience points (XP) to encourage engagement. A machine learning–based recommendation engine analyzes user performance, including accuracy, time complexity, and coding efficiency, to suggest optimized learning paths and identify weak areas. Additionally, an AI-powered code analysis module evaluates submissions and provides improvement insights. The platform includes coding tests covering multiple topics, generating detailed performance summaries to guide learners effectively. Supporting features such as real-time notifications, flash notes, admin-user chat support, and topic-based community discussions further enrich the learning experience. The proposed system aims to bridge the gap between theoretical understanding and practical coding proficiency by offering an adaptive, interactive, and scalable environment for DSA preparation.

DOI: https://zenodo.org/records/20107776

Scholaraid: Government Scholarship & Scheme Recommender

Authors: Associate Professor C.P. Lachake, Nitin Naik, Pranit Jatal, Swapnil Kharade, Vineeth Ghabak

Abstract: Government schemes and scholarships are often difficult for citizens to find because information is scattered across many official websites. SCHOLARAID is a Flask-based web application that solves this problem by bringing all scheme and scholarship details into one centralized platform. The system collects verified data through manual web scraping and stores it in a structured database for easy access. Users can search and filter schemes based on category, eligibility, state, or income level. The platform also includes an AI- powered chatbot that helps users find suitable schemes through simple, natural conversation. By making information easy to understand and access, SCHOLARAID supports Digital India and helps citizens quickly discover the benefits they are eligible for.

DOI: https://zenodo.org/records/20108450

A QR Code Attendance System And Management

Authors: Mr.S.P.Gunjal, Prashant Singh, Sudarshan Vekhande, Rupali Dhamale, Ashlesha Gadade

Abstract: This paper details the design and implementation of the QR Attendance System, a contemporary, web-based solution intended to modernize and streamline student attendance management. The system is engineered around a mobile-first, dark-mode user interface (UI) developed using React.js and TailwindCSS, ensuring optimal usability and speed via device camera scanning (html5-qrcode). The back-end, powered by Flask/Python, processes attendance in realtime, applying a key feature: configurable, time-based logic to categorize status as On Time, Late, or Very Late. Attendance records are stored in a simple, portable CSV format for easy data manipulation and reporting. This architecture prioritizes cost-effectiveness, high performance, and institutional policy adherence, presenting an efficient alternative to traditional and more complex biometric attendance methods. its scalability and performance optimization, the system prioritizes cost-effectiveness, security, and maintainability by relying on open-source technologies and lightweight architecture. Thssis approach significantly reduces infrastructure complexity while maintaining high accuracy and reliability in data processing. Comparative analysis with traditional biometric, RFID, and manual attendance methods highlights the proposed system’s superiority in terms of deployment speed, adaptability, and user engagement. Furthermore, the system supports institutional customization, enabling administrators to configure academic schedules, attendance thresholds, and reporting periods. Future enhancements may include cloud-based data storage, machine learning-driven analytics for attendance trend prediction, and integration with Learning Management Systems (LMS) for automated grading and participation tracking.

DOI: http://doi.org/

Role And Future Of Artificial Intelligence In Education: A Comprehensive Review

Authors: Swapnil Sambhaji Bhosale

Abstract: Artificial Intelligence (AI) has rapidly emerged as a transformative technology across various global sectors, with education being one of the most profoundly impacted areas. The integration of AI in education has significantly improved teaching methodologies, learning paradigms, assessment accuracy, and overall student engagement. AI-based tech-nologies, such as intelligent tutoring systems, personalized learning platforms, automated grading systems, and educational chatbots, are fundamentally reshaping traditional edu-cation models. This comprehensive review paper examines the multifaceted role of Arti-ficial Intelligence in education by deeply exploring its practical applications, pedagogical benefits, operational challenges, and long-term future opportunities. The study highlights the shift from a conventional ‘one-size-fits-all’ pedagogical approach to an individualized, data-driven framework. Furthermore, the paper investigates emerging technological trends such as Generative AI, adaptive learning ecosystems, predictive analytics, and immersive virtual learning environments in the context of higher education. Despite the immense potential, the deployment of AI in education is not without hurdles. Critical challenges in-cluding data privacy, cybersecurity, algorithmic bias, ethical concerns, infrastructural and technological limitations, and the risk of over-dependency on technology by both educa-tors and learners are critically analyzed. The paper concludes that while AI possesses the unprecedented potential to revolutionize global education systems by creating more per-sonalized, efficient, equitable, and accessible learning environments, a balanced approach combining technological innovation with ethical foresight and human-centric pedagogy is essential for sustainable integration.

DOI:

Some New Type of Contraction and Common Fixed-Point Theorem in Fuzzy Metric Space

Authors: Ms. Pooja Mishra, Farkhunda Sayyed

Abstract: The main objective of writing this research paper is to prove the Common Fixed Point Theorem in Fuzzy Metric Space with a new type of contraction

DOI: https://doi.org/10.5281/zenodo.20116890

Early Migraine Prediction Using TabNet And TabTransformer: A Comparative Deep Learning Study On Clinical Data

Authors: Mst. Sabira Mahbuba Eva, Sadia Afrin Soha

Abstract: Migraine is a prevalent neurological disorder affecting approximately one billion people worldwide, causing considerable disability and significant socioeconomic consequences. Developing effective treatment options requires accurate bracketing of migraine subtypes; yet, this process often requires specialized expertise that may be lacking in many healthcare settings. Although there are yet few comparative assessments of these infrastructures, attention-grounded deep literacy approaches for irregular clinical data suggest a potential path for automated bracket support. Using structured clinical data, this work provides a systematic relative analysis of TabNet and TabTransformer for early migraine prediction. The Kaggle Migraine Dataset, which included 100,000 case records with 24 clinical characteristics and seven migraine subtypes, was used to create the dataset. To address class imbalance, the approach used balanced class weighting, categorical encoding, and stratified train-test separation. F1-score, ROC-AUC, recall, precision, and delicacy were used to measure performance. With a test accuracy of 99.50 as opposed to TabTransformer's 92.38, experimental data show that TabNet works noticeably better than TabTransformer. Perfection (99.54), recall (99.50), F1-score (99.51), and near-perfect ROC-AUC (0.9999) were all enhanced by TabNet, with consistent performance across all seven migraine subtypes. Direct point significance visualization was made possible by TabNet's effective attention medium, providing clinically interpretable perceptivity that met ICHD-3 individual criteria. Because of its quick confluence and memory efficiency, it is useful in environments with limited resources. With significant ramifications for automated clinical decision assistance, these results validate TabNet as a model for migraine bracket tasks that is generally efficient and comprehensible.

DOI: https://doi.org/10.5281/zenodo.20117415

Digital Taxation In India

Authors: Tamakuwala Dhruvi Yash, Dr. Mehul Maheshbhai Mistry

Abstract: Digitalisation in tax has taken an important turn in tax administration and compliance in India, in the era of transparency and efficiency. Digitalisation involves the process of integrating technology and digital tools to eliminate traditional method which was labour intensive. It shows clear and precise data through automatic data entry and precise calculations. As a result of such activities, the process becomes simple, clear and less time consuming. In addition to this, chances of human error and fraud also decreases. The digitalisation of the tax system in India marks a significant shift towards greater efficiency, transparency, and compliance in public finance management. The study throws light on the development of digital era in India, in Indian tax system and in Indian economy.

High Temperature Waste Utilisation

Authors: Brijesh Kumar Yadav, Mr. Suneel Kumar Swarnkar

Abstract: High temperature waste utilisation has become one of the most important research areas in modern energy engineering and sustainable waste management systems. Rapid industrialization and urbanization have increased the quantity of industrial waste, agricultural waste, municipal solid waste, and biomass residues. Traditional waste disposal techniques such as open dumping, incineration, and landfilling create severe environmental and health-related issues. Plasma gasification technology offers an advanced solution for converting waste materials into valuable energy products. This research paper presents a comprehensive study of plasma gasification for bio-waste materials including natural waste wood, rock maple charcoal, and fiber waste wood. Experimental analysis has been carried out to investigate gasification efficiency, calorific value, syngas composition, thermal performance, and environmental impact. The study demonstrates that plasma gasification can significantly reduce harmful emissions while increasing energy recovery from biomass waste. The results indicate that natural waste wood provides maximum gasification efficiency and better syngas quality compared to other tested materials. The paper also highlights future opportunities in renewable energy generation, smart monitoring systems, and industrial waste management applications using plasma gasification technology.

A Sustainable Approach In Water Desalination Management Strategies With The Integration Of Renewable Power System

Authors: Md Amzed Hossain, Md Taher Al Masum, Md. Wahid Farhen, Effat Jahan

Abstract: This research investigates the application of solar energy in water desalination systems as a sustainable alternative to conventional fossil fuel-based methods. The proposed system integrates renewable energy with water purification to address freshwater scarcity and energy depletion challenges. Utilizing a water level sensor, the system monitors water from multiple sources, including a storage tank, deep tube well, and harvested rainwater. A reverse osmosis (RO) pump powered by solar energy filters the water, which is then stored in a main tank for safe drinking purposes. The system operates in hybrid mode—drawing power from solar energy primarily, with an automatic switch to grid power when solar availability is insufficient. Real-time data on energy availability and water status are displayed and stored in Google Sheets for monitoring and analysis. This integrated solar-powered desalination system demonstrates a cost-effective and environmentally friendly approach to sustainable water purification in remote or resource-limited areas.

DOI: https://doi.org/10.5281/zenodo.20118347

Enterprise Advanced Campus Network Security Architecture

Authors: M. Sri Krishna, M. Vishnuvardhan, R. Annamaiah, Mrs.Sivaselvi. k

Abstract: In today’s corporate landscape, the high-speed data transit and robust cybersecurity is no longer optional, it is a fundamental requirement for business continuity. This paper details the design and deployment of Enterprise Campus Network Security Architecture, using Cisco Packet Tracer. The primary objective was to build a resilient “Defense-in-Depth” security architecture for a distributed enterprise campus. The architecture is built upon a hierarchical three-tier mode, utilizing a redundant core powered by the Hot standby Router Protocol (HSRP) and OSPF dynamic routing to eliminate single point of failure. At the enterprise edge, Cisco ASA firewalls serve as the primary gatekeepers, implementing stateful packet inspection and isolating servers within a secure Demilitarized Zone (DMZ). The critical component of this project is to establish a Site-to-Site IPsec VPN tunnel, which provides an encrypted bridge between the Headquarters and the Branch network, ensuring the sensitive data remains confidential while Routing data across public ISP networks. We make the network a lot safer right at the entry point by locking down the switch hardening techniques, including Port Security witch Sticky MAC addressing, DHCP Snooping, and Dynamic ARP Inspection (DAI) to mitigate internal ‘Man-in-the-middle’ threats. The Successfully implementation of these technologies together, we’ve built a network that’s both highly reliable and easy to use, without compromising on security. It strikes a balance between letting employee access what they need and maintaining a strict ‘Zero-trust’ policy.

Allotease A University Management System

Authors: Varun Karthik Guttu , Alamooru Babaiah, Vasanta Sai Prasad, Dr N SundaraRajulu, Dr V ravindra krishna Chandar

Abstract: Allot Ease is a next-gen University Management System (UMS) designed to streamline the process of the student management till the end of his / her graduation this streamline process not only engages the student side but also it adapts and manages the entire university or a college or a school. This provides an easy management of the University that zeroing the paper work from the beginning of the student enrolment till his / her graduation that enables a smoother and safer experience with a secured data and allow them to make easier of their work to experience their arrangements. It also enables them to make the work easier even without any administration rehearses. Allot Ease manages the various sector like Exam cell, HR Administration, Academics and Finance etc., eventually managing all the directories of university from student to admin as of registrar vice chancellor that enables to flexibility

Correlating Micro-Mineralogy With Seismic Optimization & Structural Response Of Transmission Infrastructure On Geopolymer-Stabilized Subgrades

Authors: Dr. Kamalkishor M. Sharma, Shashikant R. Yadav, Monika J. Atole

Abstract: The increasing demand for electrical energy necessitates the development of economical and lightweight transmission line towers. However, constructing these lifeline structures on Black Cotton Soil (BCS) poses significant civil engineering challenges due to the soil's expansive nature and high montmorillonite content, which leads to unequal foundation settlement and low bearing capacity. This research presents a dual-phase approach: first, stabilizing the problematic BCS using Construction and Demolition (C&D) waste-based geopolymers , and second, evaluating the structural response of a 40 m transmission tower with varying bracing patterns (single, double, and knee) resting on this stabilized stratum. Utilizing STAAD.Pro software for dynamic structural analysis and advanced micro-mineralogical testing (XRD analysis) for soil evaluation , the study aims to achieve the most stable, economical tower configuration and a resilient foundation.

Role of Fuzzy Mathematics in Artificial Intelligence

Authors: Mr. Ghadage V.D.

Abstract: Artificial Intelligence (AI) has become one of the most rapidly growing fields in modern technology. AI systems often deal with uncertain, vague, and imprecise information. To handle such situations, Fuzzy Mathematics plays a crucial role in providing a mathematical framework for approximate reasoning. Unlike classical mathematics, fuzzy mathematics allows partial truth values between 0 and 1, making it suitable for real-world decision-making problems. Concepts such as fuzzy sets, membership functions, fuzzy logic, and fuzzy relations are widely used in control systems, machine learning, pattern recognition, robotics, and expert systems. This paper explores the importance of fuzzy mathematics in AI, its core components, applications, advantages, challenges, and future scope. The study highlights how fuzzy-based approaches improve decision-making, handle uncertainty, and enhance intelligent system performance.

DOI: https://doi.org/10.5281/zenodo.20132633

AI-Enabled Virtual Health Check-Up and Diagnosis Support System

Authors: K. Venkat Naveen, Mandati Santosh, Matta Rahul 3, Mr.K.Saravanan

Abstract: AI-driven healthcare diagnostic systems are a critical component of modern medical data analysis, as they directly influence diagnostic accuracy, clinical efficiency, and patient safety. Their primary objective is to assist healthcare professionals by analyzing large volumes of medical data while enabling faster and more reliable clinical decision-making. This paper discusses current technical approaches to AI-assisted healthcare systems, focusing on the application of artificial intelligence, machine learning, and deep learning techniques to process medical images, patient records, and clinical datasets. The relationship between intelligent diagnostic systems and predictive healthcare models is also examined, highlighting their growing importance in next-generation medical technologies. Limitations in traditional diagnostic processes and manual data analysis methods are identified based on observations from recent healthcare practices. These findings emphasize the need for advanced AI-based solutions that enhance diagnostic accuracy, efficiency, and reliability while supporting existing medical workflows rather than replacing healthcare professionals.

E-Governance Based Gram Panchayat Management System Using Blockchain Technology

Authors: Prof. S. P. Gunjal, Sakshi Takhik, Shraddha Totre, Pratiksha Pawar, Shubham Ahire

Abstract: The E-Governance Based Gram Panchayat Management System Using Blockchain Technology aims to enhance transparency, security, and efficiency in rural governance by digitizing Panchayat operations. Traditional systems often face challenges such as data manipulation, lack of accountability, and delays in service delivery. This proposed system integrates blockchain technology to provide a decentralized and tamper-proof platform for managing records, transactions, and citizen services. It enables functionalities such as certificate issuance, grievance handling, and scheme management through a secure and user-friendly interface. Smart contracts are used to automate processes, ensuring faster and reliable service delivery. Overall, the system improves trust, reduces corruption, and promotes effective and transparent administration at the Gram Panchayat level.

Intelligent Poultry Farm Management System Using Machine Learning and IoT: Architecture, Implementation, and Performance Evaluation

Authors: Gauri Vijay Shinde, Aarti Santosh Shinde, Manisha Rajendra Shinde, Dr. Shubhangi Gunjal, Professor Bangar A.P, Professor Bhosale S.B.

Abstract: Poultry farms traditionally rely on non-digital, reactive monitoring practices. This results in a delay in detecting unfavourable environmental conditions. In turn, this leads to increased bird deaths and economic loss. The Intelligent Poultry Farm Management System (IPFMS) is a project that has been designed for the commercial broiler farm that is an IOT and machine learning based project. The system utilizes five types of sensors which are MQ135 (ammonia), DHT11/BME280 (temperature and humidity from the feed presentation), load cell (feed weight), and HC-SR04 (water level). So, these sensors connect to the ESP32 microcontroller and transmit real-time data. Moreover, it connects to MQTT Server (broker) which is an instance/service hosted in the cloud. So, the service ultimately provides an Apache/MySQL back-end. Three ML algorithms are employed for distinct prediction tasks: to forecast the ammonia levels, we use linear regression; for classifying risks and safety, we use decision tree; and logistic regression is used to compute the probabilities of the various risks. The farmers are notified for each tier (Normal, Warning, Critical) through a real-time web & mobile dashboard. The end-to-end latency is less than 1.5 seconds. A connected e-commerce module allows farm level transactions. Tests conducted on a live commercial broiler farm with 5000 birds, over 30 days found the sensor visible 99.7% of the time, Zero Critical false alerts, 92.9% Warning Tier Correct alert in a Commercial biosecurity farm set-up. The comparative study shows that IPFMS achieves better accuracies, alerts, and functionalities than existing IoT-only and single-model approaches, thereby providing a real-life and scalable blueprint for smart poultry farming 2.0.

DOI: https://doi.org/10.5281/zenodo.20135121

IoT-Based Smart Drip Irrigation System with Real-Time Soil Moisture Monitoring and Weather-Aware Predictive Control

Authors: Bhosale Srushti G, Dongare Shweta S, Kabadi Savali S, Professor Bangar A. P, Professor Bhosale S. B

Abstract: Water shortage and inefficient forms of irrigation are the two most pressing challenges facing world agriculture today, especially in less developed countries. Numerous farmers still utilize manual irrigation or fixed-time based scheduling that leads to excessive water usage, uneven crop growth, and economic loss in many cases. To solve these problems, this work offers the design, implementation and experimental validation of an IoT Based Smart Drip Irrigation System (ISDIS) which automatically controls irrigation by monitoring real-time soil moisture and ambient weather condition. Soil moisture sensors used in the system are the capacitive type and connected to the DHT11/DHT22 temperature-humidity sensor via a NodeMCU/ESP32 microcontroller. Data from the sensors is uploaded over encrypted Wi-Fi after every 10 seconds to the Firebase Realtime Database. Farmers can monitor their field conditions remotely using an Android application developed in Java/XML. The valve actuation is governed by a dual-criteria decision algorithm. (i) Irrigation is activated when the soil volumetric water content (VWC) is below a calibrated threshold, and (ii) Irrigation is proactively deferred when a rain forecast for greater than 5 mm occurs within a 12-hour window based on data from the OpenWeatherMap API. Through empirical validation over a cultivation cycle of 90 days, the system was able to achieve an overall water saving of 35.02%, the soil VWC stability was enhanced by 66.49% an average actuation delay of 2.1 s, and average crop-yield increase of 19.5% over five crops when compared to the Fixed-Schedule Irrigation (FSI). Through the timely deferral of 14 irrigation cycles based on meteorological forecasting, the system was successfully realized at a hardware cost of approximately USD 30, providing a low-cost scalable energy-efficient framework for precision agriculture and sustainable smart farming.

DOI: https://doi.org/10.5281/zenodo.20135762

SafeURL AI Extension: An Intelligent Browser Extension for Real-Time Detection of Malicious and Phishing Content

Authors: Dhumal Pratiksha Shivnath, Thorat sakshi Sachin, Kale Omkar Santosh, Professor Bangar Abhishek Popat

Abstract: The rapid growth of cyber threats such as phishing websites, malicious scripts, fake academic journals, and deepfake videos has created serious challenges for online security and trust. Traditional protection systems based on static blacklists and signature detection are no longer sufficient to handle evolving and zero-day attacks. This paper proposes SafeURL AI Extension, an intelligent browser extension that integrates machine learning and real-time analysis to detect unsafe digital content and protect users during web browsing. The system incorporates three tightly integrated modules: (i) malicious URL detection using Logistic Regression and Random Forest classifiers trained on structural URL features and webpage behavioral signals; (ii) academic journal authenticity verification through DOI resolution, ISSN cross-checking, DOAJ indexing validation, and domain-blacklist matching; and (iii) deepfake video analysis using a CNN-based deep learning pipeline that processes frame-level facial artifacts to classify content as real or manipulated. Evaluation results demonstrate that the URL detection module achieves approximately 95% accuracy, while the deepfake module achieves 99.9% confidence on manipulated samples. The system provides risk scores, color-coded warnings, and actionable user alerts within 1–2 seconds of page load. By combining real-time threat detection, adaptive learning, and privacy-preserving local analysis, SafeURL AI Extension offers an effective, scalable, and lightweight solution for enhancing browser-level cybersecurity and ensuring a safer digital environment for everyday users.

DOI: https://doi.org/10.5281/zenodo.20136248

Classical Contextuality And The Limits Of Bell’S Theorem: A Vector‑Based Analysis

Authors: Rakin Khan

Abstract: While Bell’s theorem is traditionally seen as a total rejection of local classical models in quantum mechanics, this paper argues that such a conclusion relies on the narrow assumption of "joint definiteness." Bell’s derivation requires a single hidden state, λ, to account for all possible measurement outcomes simultaneously, which implies that the measurement apparatus remains microscopically unchanged regardless of its settings. However, because physical instruments are composed of matter, their internal microstates necessarily vary with the chosen settings, meaning the relevant hidden variables are setting-dependent. This physical reality prevents the formation of a global joint probability distribution, placing these types of models outside the mathematical constraints of Bell’s and Fine’s theorems. Consequently, the authors conclude that while Bell’s theorem remains mathematically sound, it only excludes hidden-variable models that ignore the setting-dependent nature of the measuring apparatus, leaving the door open for classical descriptions of physical reality.

DOI: https://doi.org/10.5281/zenodo.20136490

A Comparative Evaluation Of The Physicochemical Characteristics Of Shivganga Pond And Nandan Pahar Pond Was Conducted To Investigate The Influence Of Anthropogenic Activities On The Variation And Deterioration Of Water Quality Parameters.

Authors: Pritam Sharma, Dr. Nilesh Kumar, Dr. N.K. Mandal

Abstract: The present investigation is centered on the limnological evaluation of Shivganga Pond, an ecologically and culturally important water body situated near the renowned Baidyanath Temple, along with Nandan Pahar Pond, which serves as an important source of drinking water for the residents of Deoghar through an established water distribution system. The study involves a comparative analysis of sixteen physicochemical parameters of both ponds, including temperature, pH, dissolved oxygen (DO), chemical oxygen demand (COD), and other related water quality indicators, in order to evaluate the ecological condition, water quality status, and overall sustainability of these aquatic ecosystems. The observed differences in water quality characteristics between the two urban ponds reflect varying levels of anthropogenic pressure and clearly demonstrate the influence of human activities on the deterioration and modification of aquatic environments. The variations recorded in the physicochemical parameters of these water bodies, when compared with the standard limits recommended by the World Health Organization, highlight the growing need for both scientific understanding and public awareness regarding the protection and conservation of freshwater resources. The findings of the study underline the importance of adopting sustainable, integrated, and community-oriented water management strategies to preserve the ecological balance of these ponds. Furthermore, the study aligns with the objectives of the United Nations Sustainable Development Goals (SDGs), particularly SDG 6, which focuses on clean water and sanitation, and SDG 14, which emphasizes the conservation and sustainable use of aquatic ecosystems and life below water.

DOI:

RMM Abuse Detection and Prevention Research Report Cyber Security Research Project Topic: Detection and Prevention of Remote Monitoring and Management (RMM) Abuse

Authors: Sujeet Gautam

Abstract: Remote Monitoring and Management (RMM) tools have become an important part of modern enterprise IT infrastructure. These tools are widely used by Managed Service Providers (MSPs), system administrators, and IT support teams to remotely manage devices, monitor infrastructure health, deploy software updates, and troubleshoot technical issues. However, cybercriminals increasingly exploit legitimate RMM software to gain stealthy and persistent access to victim systems without using traditional malware. This report explores the growing threat of malicious RMM abuse, analyzes existing detection approaches, and proposes a hybrid detection framework capable of identifying suspicious RMM behavior through machine learning, behavioral analytics, and network fingerprinting techniques. The study also evaluates current challenges, discusses attacker evasion strategies, and provides recommendations for vendors and defenders to improve security posture. The research concludes that hybrid behavioral analysis combined with telemetry correlation significantly improves detection accuracy while reducing false positive rates compared to traditional signature-based detection systems.

DOI: https://doi.org/10.5281/zenodo.20137033

AI Interview Platform: Intelligent Android-Based System for Automated Interview Practice and Evaluation

Authors: Mitali Vadiya, Dnyanal Pawar, Akshada Kolhe, Dr.A.A.Khatri, Professor Bangar.A.P, Profesor Bhosale S.B

Abstract: The AI Interview Platform is an Android-based system that uses Artificial Intelligence (AI) and Natural Language Processing (NLP) to simulate real interview scenarios and provide personalized feedback. Developed using Java/XML with Firebase Realtime Database, it allows users to practice interviews through a virtual interviewer by submitting spoken or typed responses. The system evaluates communication aspects such as clarity, grammar, vocabulary, tone, confidence, and technical accuracy. A Generative AI engine offers instant suggestions to improve response structure and communication skills. The platform also includes quiz-based assessments and video learning modules to enhance both technical and soft skills. With adaptive practice and cloud-based analytics, it provides a scalable solution for improving interview readiness.

DOI: http://doi.org/

Jarvis AI: A Python-Based Personal Voice Assistant

Authors: Surendra Mahendra Kamble, Atharv Ganesh Hande, Sumit Vaman Ghatul, Professor Abhishek, Popat Bangar

Abstract: Artificial intelligence (AI) has drastically changed the way humans and computers interact. Deep learning architectures particularly transformer-based model have played a major role in revolutionising spoken language understanding, intent classification and generative response generation. The paper designs, implements and evaluates JARVIS AI, a personal voice assistant written in Python using a range of advanced AI techniques. The system is composed of various intelligent modules including speech recognition, natural language processing (NLP), desktop automation, real-time web search, and AI-based image generation. In the quiet condition, the performance of the Proposed system works very good. It achieves 94.7% speech recognition accuracy. The average response time for command is less than 500 ms and 96.1% successful task execution rate with 10 functionalities [1,2]. JARVIS AI has a modular architecture that allows for easy scalability, flexibility, and integration of new features JARVIS AI enables robust offline use and with deep-software automation unlike commercial cloud-based systems. The findings of the experiments show that JARVIS can be an efficient, robust and user-friendly digital assistant who can enhance one’s personal and professional productivity significantly.

DOI: https://doi.org/10.5281/zenodo.20139140

Multimodal Machine Learning for Image-Based Clinical Diagnostics in Biomedicine: A Result-Driven Comparative Study

Authors: Dr. A. A. Khatri, Professor V. A. Karad, Shreyash S. Andre, Akash S. Mundhe, Ashish C. Nalawade, Professor Bangar A.P

Abstract: The rapid increase in imaging data, electronic health records, lab reports, and patient-doctor discussions has created an important need for intelligent systems to understand and interpret disparate biomedical data in a unified manner. The conventional unimodal diagnostic systems that solely rely on textual reports or medical images do not capture the complete clinical context for adequate decision-making. The study explores the application of multimodal machine learning (MML) to solve clinical diagnostic problems using images. Here we describe a health chatbot that uses the Gemini API by Google. It also outlines an X-ray and medical-image analyzer based on convolutional neural network (CNN). An audio-based health assistant helps to transcribe doctor-patient conversations and summarize them. The combined report generator joins heterogeneous PDFs into a single intelligible document. The proposed components of the MML framework, and their fusion integration, are demonstrated in a set of experiments. Human trials produced diagnostics that are indistinguishable from those by human clinicians. The framework suggested was implemented using Python, Streamlit, PyPDF2, ReportLab, Pandas, and Matplotlib Google Gemini Pro Vision API. This was assessed on a well-crafted dataset with 1,000 chest X-ray images belonging to 5 diagnostic classes, i.e., Normal, Pneumonia, Tuberculosis, COVID-19, and Lung Cancer, 500 corresponding clinical reports, and 200 audio consultations. The experimental outcomes show that the multi-modal system has an overall accuracy of 94.6%, with a precision of 93.8%, recall of 94.1%, F1-score of 93.9% and AUC of 0.96. These values are significantly higher than those of the unimodal baseline SVM (78.4%), The results show that the multimodal fusion of text, image and audio modalities, along with explainable visualization, dramatically improves diagnostic accuracy, clinical workflow efficiency and patient understanding.

DOI: https://doi.org/10.5281/zenodo.20139048

AI-Based Acoustic Wave Monitoring of Rail Defects and Predictive Rail Wear Analysis

Authors: Asawari Gawade, Pranita Jadhav, Dhananjay Rode, Prof.Bangar A.P, Prof.Bhosale S.B.

Abstract: Indian Railways is one of the largest rail networks in the world. The structure of the rail track is crucial for passenger safety and freight reliability. Traditional inspection techniques involving visual surveys, ultrasonic flaw detectors and hammer tests are labor-intensive, episodic and miss early-stage degradation. This study presents a comprehensive AI-based framework which involves monitoring acoustic waves through the use of real-time stress-wave acquisition which leads to defect classification with the usage of deep-learning and predictive rail-wear analysis. A mobile inspection unit that is built using an ESP32 microcontroller continuously samples vibration signals from piezoelectric and MEMS-based at 8 kHz. Moreover, it performs on-edge pre-processing of that data (which include pre-emphasis filtering, Hamming windowing, FFT and Mel-Frequency Cepstral Coefficient extraction). After that, it transmits feature vectors to the cloud server through Wi-Fi using the REST API. A CNN+LSTM scheme classifies the defect into normal, surface crack, internal flaw, joint defect and severe wear. The stacked-LSTM regressor forecasts cumulative wear at a 180-day horizon and uses historical sensor data along with outdoor temperature and traffic load data for its predictions. Field trials conducted on a 2.4 km test track produced 12,500 labelled samples. The proposed model achieved an overall classification accuracy of 97.2 %, precision of 96.5 %, recall of 96.9 % and F1-score of 96.7 %. The model outperforms other benchmarks which includes SVM (86.3 %), Random Forest (89.7 %), KNN (82.1 %), CNN-only (93.4 %) and LSTM-only (94.6 %). The wear-prediction module achieved a Mean Absolute Error (MAE) of 0.18 mm in comparison of 0.61 mm for ARIMA and 0.84 mm for linear regression. A web dashboard made using HTML, CSS and JavaScript visualises live track condition, GPS coordinates, vibration spectra, alert logs and wear-forecast curves. The system facilitates condition-based predictive maintenance, reduces the life cycle cost by approximately 28% and presents low-power, scalable solution for next generation railway-health monitoring while reducing dependency on manual inspection.

DOI: https://doi.org/10.5281/zenodo.20138968

Non Invasive Method Of Varicose Vein Detection Using Raspberry Pi

Authors: Keerthika.S, Sharmi.S, Yuvasree.S, Mrs.K.Deepa, Deepa.K

Abstract: Visualization of subcutaneous veins is essential for procedures such as venipuncture and intravenous injections. Conventional methods basedonvisual inspection andpalpationare often unreliable, particularly in patients with lowvenous visibility, leading to multipleinsertion attempts and discomfort.Thispaperproposesanon-invasivenear-infrared(NIR)imaging system for real-time vein visualization. The method exploits the higher absorption of infrared radiation by deoxygenated blood, allowing veins to appear with enhanced contrast. An infrared-sensitive camera captures images, which are processed using grayscale conversion, contrast enhancement, and edge detection techniques. The proposed system is low-cost, portable, and user-friendly, making it suitable for clinical assistance and preliminary venous assessment. Experimental results demonstrate the effectiveness of NIR imaging for accurate vein detection..

DOI: https://doi.org/10.5281/zenodo.20151385

A Hybrid Context-Aware Hinglish Spam And Scam Detection System Using Ensemble Machine Learning

Authors: Siddhi Gupta, Ms. Dimpy Parashar, Dr. Yatu Rani

Abstract: The rapid expansion of digital messaging platforms has given rise to an alarming surge in spam and scam communications, particularly in linguistically diverse regions like India where users frequently communicate in Hinglish—a dynamic blend of Hindi and English. Traditional spam detection systems are designed exclusively for monolingual text and fail to process such code-mixed linguistic patterns effectively. This paper presents a hybrid, context-aware spam and scam detection framework that combines rule-based keyword filtering with ensemble machine learning techniques. The system performs multi-stage text preprocessing, applies TF-IDF feature extraction, and trains three individual classifiers—Logistic Regression, Naive Bayes, and Support Vector Machine—fused through soft-voting ensemble learning. Evaluated on a curated dataset of 5,200 Hinglish and English messages, the proposed Voting Classifier achieves an overall accuracy of 94.7%, precision of 93.8%, recall of 94.5%, and F1-score of 94.1%, surpassing all individual baseline models. The results confirm that ensemble fusion combined with domain-specific rule-based filtering meaningfully improves detection coverage for code-mixed, real-world spam and scam messages.

DOI: https://doi.org/10.5281/zenodo.20153347

A Unified Information-Theoretic Model of Cosmological Cycles and the Self-Optimization Imperative: The Universal Substrate, Emergent Reality, and Dual Recursive Processing

Authors: Swaminathan Mani

Abstract: Current Theoretical Synthesis: This paper introduces the Universal Substrate (US), a discrete, non-local information-processing architecture that serves as the ontological basis for the physical universe. This model proposes that the observable cosmos is an Emergent User Interface (UI), where the laws of physics are not fundamental constants but identified as algorithmic protocols optimized for systemic stability. By reinterpreting spacetime as a Topological Information-Braiding manifold, this model provides a unified resolution – reconciling the discrete nature of Quantum Mechanics with the geometric curvature of General Relativity through a single, self-correcting Autodidactic Meta-Algorithm.

DOI: http://doi.org/

A Lightweight Hybrid HOG–MobileNetV2–ACK-ELM Framework For Real-Time Facial Recognition

Authors: Kanupriya Upadhyay, Manisha Sharma

Abstract: In settings where resources are limited, traditional automated attendance systems have a lot of trouble because they are slow to respond and are sensitive to changes in the environment. Li et al. research presents an enhanced facial recognition framework that mitigates the efficiency deficit in lightweight segmentation as identified [1] . Ryando et al. pointed out To improve security against the spoofing vulnerabilities [2] in healthcare kiosks, we add a liveness detection module that is based on the Eye Aspect Ratio (EAR) looked at heavy deep learning architectures that put a lot of strain on computers. In contrast, our system uses an Adaptive Circular Kernel Extreme Learning Machine (ACK-ELM) to achieve a non-iterative, one- shot learning paradigm. This method is based on the fast hybrid ideas of Anil and Suresh [4] and combines Histogram of Oriented Gradients (HOG) with shallow MobileNetV2 features. To ensure resilience against the lighting and expression varia- tions examined by Abdallah et al., the model is validated on the Extended Yale B dataset. The practical deployment phase comes after the five-phase attendance marking architecture that Potdar et al. [6] proposed. It uses a MySQL database to keep track of records in real time. Our classification strategy also works better than the Support Vector Machine (SVM) baselines set by Ali et al.[8] because it makes inferences faster. By using the few-shot learning efficiencies that Nasralla [9] looked into in the AIFS framework, the system stays very accurate even when it doesn’t have a lot of training data. To make the best use of memory, we use the Global Average Pooling (GAP) techniques suggested by Wei et al. [9]. These techniques compress features to keep the system from crashing. Lastly, the system uses the multimodal fusion logic of Abdul-Al et al. [10] and the temporal consistency principles of Interno` et al. [11] to tell the difference between real facial trajectories and fake or static ones. Experimental results show that the accuracy is 97.16% on a standard CPU, which makes it a useful solution for large-scale institutional attendance.

DOI: http://doi.org/10.5281/zenodo.20156258

Cyber Security Governance And Legal Challenges In Secure Cloud Architecture

Authors: Mr. Shantanu, Dr. Tanu Arora, Dr. Narinder Khubber

Abstract: This paper will examine the evolving cyber security threats associated with cloud computing and will analyse the legal implications arising from inadequate data protection mechanisms and insecure cloud architectures. It will further evaluate the role of encryption standards, secure-by-design cloud frameworks, identity and access management systems, and risk mitigation strategies in ensuring legal compliance and cyber resilience. Special emphasis will be placed on international and national legal frameworks governing cloud security, including the European Union’s General Data Protection Regulation (GDPR), the Health Insurance Portability and Accountability Act (HIPAA), the Payment Card Industry Data Security Standard (PCI DSS), and India’s Digital Personal Data Protection (DPDP) Act. The study will also explore issues relating to jurisdiction, data sovereignty, liability of cloud service providers, contractual obligations, cybercrime investigation, and regulatory enforcement in cloud ecosystems. Furthermore, the paper will propose compliance-oriented and legally sustainable cyber security strategies that organizations will be expected to adopt in future cloud infrastructures. The research will conclude that effective cloud governance will require the integration of technological safeguards with strong legal and regulatory frameworks to ensure data privacy, accountability, and secure digital transformation.

DOI: http://doi.org/10.5281/zenodo.20156680

Seismic Performance Evaluation Of Irregular RCC And Steel Buildings Under Different Soil Conditions Using STAAD.Pro

Authors: Ambuj kumar Yadav

Abstract: The seismic performance of buildings is strongly governed by structural irregularities and soil characteristics. This study conducts a comparative analysis of irregular RCC and Steel buildings using STAAD.Pro under three soil conditions defined in IS 1893:2016—Hard, Medium, and Soft. Plan irregularity (L-shape), vertical irregularity (setback), and mass irregularity are incorporated to assess their influence on dynamic behaviour. The Response Spectrum Method is adopted to evaluate key seismic parameters, including storey displacement, drift, base shear, time period, and mode shapes. Results indicate that soft soil significantly amplifies seismic demand, leading to larger lateral responses. Steel buildings display higher flexibility with lower base shear, while RCC structures exhibit greater stiffness and reduced displacement. Irregular configurations also exhibit notable torsional effects.

Design And Analysis Of Retaining Wall Load Force And IRC Bridge Loading Using Software Auto CAD & STAAD Pro.

Authors: Neeraj Kumar Gautam, Er. Daljeet Pal Singh

Abstract: This study presents a comprehensive methodology for the design and analysis of bridge retaining walls and abutments subjected to Indian Roads Congress (IRC) bridge loading standards, specifically focusing on IRC:6. The structural integrity of these components is critical for the overall stability of bridge systems, necessitating precise load estimation and modeling. The research integrates computational tools, utilizing STAAD.Pro for finite element modeling and structural analysis, and AutoCAD for detailed engineering drafting and reinforcement layout.The analysis workflow begins with the derivation of superstructure reactions (dead, live, and impact loads) according to IRC:6 provisions, coupled with geotechnical lateral earth pressures. Seismic effects are evaluated using the Mononobe-Okabe analytical method, supplemented by finite element verification to account for non-linear soil- structure interaction. In STAAD.Pro, the wall and connected wing walls are modeled using four-noded plate elements to capture complex bending, shear, and torsional behaviors. Results from the analysis—including moment and shear envelopes—are used to optimize reinforcement detailing. The study concludes with the generation of high-precision construction drawings in AutoCAD, ensuring adherence to code-specified anchorage and drainage requirements. This integrated approach ensures a robust, code-compliant design that balances structural safety with computational efficiency.

Railway Track Safety Using VSWR Based Fault Detection System

Authors: Deepti Gore, Shravani Bagayatkar, Samiksha Chopade, S. M. Hambarde

Abstract: Railway infrastructure is critical, and track faults pose significant safety and operational risks. This research presents a novel, real-time Voltage Standing Wave Ratio (VSWR)-based fault detection system for continuous railway track integrity monitoring. The system leverages the principles of Time-Domain Reflectometry (TDR), where changes in the rail's characteristic electrical impedance due to faults (like cracks, rail breaks, or joint degradation) are detected as significant spikes in the measured VSWR. A dedicated hardware module continuously injects a signal and monitors the reflected energy. A Machine Learning (ML) classifier (e.g., Support Vector Machine or Random Forest) is then employed to analyze the VSWR waveform, categorize the fault type, and accurately estimate its location. This approach offers a non-intrusive, continuous, and low-latency alternative to traditional, periodic inspection methods. The proposed system aims to significantly enhance railway safety, minimize operational maintenance costs, and prevent catastrophic failures.

Constraint-Based Intelligent Exam Seating Optimization System

Authors: Hemaganesh.B, Hemanth.R, I.Srinivas

Abstract: The Constraint-Based Intelligent Exam Seating Optimization System is developed to automate the process of assigning seats to students during examinations in an efficient and organized manner. In many educational institutions, exam seating arrangements are prepared manually, which can be time-consuming and may lead to errors, especially when handling a large number of students. This project aims to simplify and optimize the seating arrangement process by using constraint-based techniques that ensure proper seat allocation while following specific rules and conditions. The system considers important constraints such as preventing students of the same subject from sitting next to each other, maintaining classroom capacity, and ensuring a fair distribution of students across available examination halls. By processing student information and available room data, the system generates an optimized seating arrangement quickly and accurately. Additionally, the system includes notification features that inform students about their exam seat number and room details through email and SMS alerts. This reduces confusion and improves communication between administrators and students. Overall, the proposed system helps educational institutions manage examinations more effectively by reducing manual effort, minimizing errors, and improving the overall efficiency of exam seating management.

DOI: https://doi.org/10.5281/zenodo.20176113

Animal Intrusion Detection System With AI Automation: A Solar-Powered IoT Framework Combining Radar, Thermal Sensing, And YOLOv8 Deep Learning For Real-Time Wildlife Monitoring

Authors: Gabhale Ankita Dattatray, Lawande Sayali Prabhakar, Mahabare Pratham Ravindra, Prof. A. P. Banger, Prof. K. D. Dere

Abstract: Human–wildlife conflict in forest-border regions has emerged as a critical concern for both rural communities and conservation agencies, with increasing incidents of large carnivores such as tigers, leopards, and hyenas entering human settlements. Conventional surveillance methods—watchtowers, perimeter fencing, and passive CCTV monitoring—rely heavily on continuous human observation and are therefore prone to delayed responses, fatigue-induced errors, and high operational cost. This paper presents the design, implementation, and experimental evaluation of an AI powered Animal Intrusion Detection System (AniDet) that integrates Internet of Things (IoT) hardware, embedded processing, and a deep-learning vision pipeline. The prototype employs six RCWL-0516 Doppler radar modules to provide 360° motion coverage, an MLX90614 non-contact infrared (IR) thermometer to confirm the presence of a warm body, and an ESP32- CAM module that streams JPEG frames over Wi-Fi/GSM to a Flask backend running a YOLOv8m classifier trained on a custom three-class dataset (tiger, leopard, hyena). On a held-out test set of 200 frames, the system achieved 92.0 % overall classification accuracy, mAP@0.5 of 0.93, and a mean detection confidence of 0.81 across the three target classes. End-to-end latency from motion to alert was measured at ≈ 1.68 s, while the multi-stage sensor fusion (radar + thermal + vision) reduced false alarms from 38 % (radar-only) to 6 %. The complete system operates on a 6 W solar–TP4056–Li-ion subsystem, demonstrating its suitability for off-grid deployment in forest-fringe regions.

DOI: http://doi.org/10.5281/zenodo.20176466

“Ai-Powered Handheld Device For Malaria And Anemia Detection”

Authors: Vinod, Prajwal A R, Kasturi Shantesh Halasagi, Meghana L D, Pradeep Kumara V H

Abstract: This project presents a low-cost, AI-powered handheld device for detecting Malaria (Plasmodium vivax) and Anemia from Giemsa-stained blood smear slides. The system uses an ESP32-CAM module for edge-based inference via TensorFlow Lite for Microcontrollers (TFLite Micro), eliminating the need for cloud connectivity or laboratory infrastructure. A MobileNetV2 CNN model quantized to INT8 precision is embedded on-device to classify blood smear images and output confidence scores. Results are displayed on a 1.8-inch SPI TFT screen and logged to a 32GBmicroSDcard. The device is powered by a5000mAh USB power bank and is built entirely from affordable, off-the-shelf components. It is designed for use by community health workers in rural and resource-constrained settings, providing a portable, battery operated, point-of-care diagnostic tool that does not depend on internet access, trained laboratory staff, or expensive equipment.

DOI: https://doi.org/10.5281/zenodo.20176874

Controlled Self Evolving Code Optimization Engine

Authors: K. Deepthi, M. Janaki Reddy, M. Prathyusha, Dr. J. Mercy Geraldine

Abstract: This Software systems require continuous optimization to maintain performance, scalability, and maintainability. Traditional optimization approaches rely on manual developer intervention or static compiler-level improvements that cannot adapt to runtime variations. Recent AI based programming tools provide automated suggestions but lack autonomous validation and controlled evolution capabilities [1],[2]. This project proposes a Controlled Self-Evolving Code Optimization Engine, an intelligent system that analyzes execution behavior, detects inefficiencies, generates opti mized code variants, validates them through automated testing, and selectively integrates improvements. The proposed system introduces controlled software evolution that ensures safety, reliability, and adaptive performance tuning while maintaining developer oversight.

DOI: https://doi.org/10.5281/zenodo.20177151

Topological Data Analysis Driven Multi Resolution Sinusoidal Machine Learning Model For Early Detection Of Parkinson Disease Using Gait And Sensor Data

Authors: Dhanush G, Stalin Rayappan S

Abstract: – Parkinson’s Disease (PD) is a progressive neurological disorder that mainly affects motor functions such as speech, movement, and coordination. Early detection plays a crucial role in improving patient care and slowing disease progression. In recent years, machine learning techniques using biomedical signals have shown promising results for non-invasive diagnosis. This paper presents a novel approach based on a Topological Data Analysis (TDA)-driven multi- resolution sinusoidal machine learning model for early detection of Parkinson’s Disease using voice, gait, and sensor data. The proposed system follows a complete data processing pipeline that includes data preprocessing, feature scaling, feature selection using mutual information, and dimensionality reduction using Principal Component Analysis (PCA). Classification is performed using Random Forest, Support Vector Machine (SVM), and Logistic Regression models. In addition to conventional features, the system integrates TDA to capture structural patterns in data and multi-resolution sinusoidal analysis to extract frequency-based characteristics from gait signals. Experimental analysis demonstrates that the combination of advanced feature extraction techniques with machine learning improves classification accuracy and robustness. Among the evaluated models, ensemble- based methods show better performance in handling complex biomedical data. The proposed approach provides a reliable, cost-effective, and non-invasive solution for early-stage Parkinson’s detection, making it suitable for real-world healthcare applications.

DOI: https://doi.org/10.5281/zenodo.20177653

Machine Learning-Based Human Resource Analytics Framework For Employee Attrition Prediction Using Random Forest Model

Authors: Dr. Jermiah Anand Jupalli, Dr. Kiran Koduru

Abstract: The field of Human Resource Management (HRM) has experienced a significant shift in its landscape as Artificial Intelligence (AI) and Machine Learning (ML) technologies have been introduced.The Human Resource Management (HRM) landscape is undergoing significant transformation, with the introduction of Artificial Intelligence (AI) and Machine Learning (ML) technologies. Attrition becomes one of the key challenges in the organizations as it directly affects their productivity, operational efficiency and cost. Employee turnover is influenced by a complex interdependency between employee behavioral and organizational factors that makes it difficult to accurately predict employee turnover with traditional HR management approaches. The paper suggests a Machine Learning-Based Human Resource Analytics Framework based on Random Forest algorithm for intelligent prediction of employee attrition. The proposed framework is based on employee-related attributes including job satisfaction, monthly income, work experience, working overtime, performance rating, and work-life balance that could be used to classify employees likely to quit from the organization. Experiments are performed using the IBM HR Analytics Employee Attrition dataset. The performance of the suggested RF model is compared with Decision Tree, Logistic Regression and Support Vector Machine (SVM) models. Experimental results show that the proposed Random Forest model can accurately predict the disease and the obtained accuracy of 96.8% is better than other models in terms of precision, recall and f1 score. The suggested framework helps HR units in taking strategic decisions regarding employee retention and workforce management.

DOI: http://doi.org/10.5281/zenodo.20178703

“Ai Based Predictive Maintenance And Anomaly Detection For Fleet Vehicles”

Authors: Narasimha Nayaka M S, Nirmitha H R, Rohini B R, Suguna G H

Abstract: This project presents a low-cost AI and IoT-based predictive maintenance system for fleet vehicles using ESP32 and embedded machine learning techniques. The system continuously monitors important vehicle parameters such as temperature, vibration, speed, and GPS location using multiple sensors integrated with the ESP32 microcontroller. Sensor data is processed in real time and analyzed using a machine learning model developed with Edge Impulse Studio for anomaly detection and fault prediction. The system uses the Blynk IoT platform for cloud-based monitoring, live data visualization, and alert notifications. A Hall Effect sensor is used for speed detection, while the MPU6050 sensor measures abnormal vibrations in the vehicle system. The GPS module enables real-time vehicle tracking and location monitoring. During abnormal conditions, the system automatically activates buzzer and LED alerts and controls the motor through relay and MOSFET circuits for safety protection. The proposed system reduces unexpected breakdowns, minimizes maintenance costs, and improves operational efficiency and vehicle safety. The project is developed using affordable and easily available hardware components, making it suitable for smart transportation and industrial monitoring applications. The developed prototype demonstrates reliable real-time monitoring, intelligent fault detection, and remote fleet management capabilities.

DOI: https://doi.org/10.5281/zenodo.20178811

Campus 360: End-To-End Education ERP

Authors: Mrs.R. Mano Ranjani, K.Vamsi, M.Stalin Babu, M. Koteswara rao

Abstract: The rapid growth of digital technologies has significantly transformed various sectors, including education. Higher education institutions require efficient, centralized systems to manage academic and administrative operations. This paper presents Campus 360, an end-to-end Education Enterprise Resource Planning (ERP) system designed to automate and integrate key institutional processes such as admissions, attendance, examinations, fee management, library services, and placements. The system utilizes a three-tier architecture with a modern web-based technology stack to ensure scalability, security, and performance. By centralizing data and enabling real-time access, Campus 360 enhances operational efficiency, reduces manual errors, and improves decision-making. Experimental deployment in a simulated institutional environment demonstrates improved workflow management and data accuracy. The proposed system provides a scalable and adaptable framework for digital transformation in higher education institutions.

DOI: https://doi.org/10.5281/zenodo.20179248

Breast Scan AI: A Weighted Soft Voting Ensemble for High-Accuracy Breast Cancer Detection Using FNA Cytological Feature Analysis

Authors: Sneha, Mr. Yug Lohchab, Dr. Akhilesh Das Gupta, Guru Gobind Singh

Abstract: Breast cancer remains the most prevalent malignancy among women globally, with approximately 2.3 million new diagnoses annually. Early and accurate automated detection is clinically critical. This paper proposes Breast Scan AI, a novel Weighted Soft Voting Ensemble (WSVE) integrating five heterogeneous base classifiers: Random Forest (RF), Extra Trees (ET), Support Vector Machine with RBF kernel (SVM-RBF), Logistic Regression (LR), and Gradient Boosting (GB). The proposed model is evaluated on the Wisconsin Breast Cancer Dataset (WBCD, UCI) comprising 569 instances and 30 cytological features. The ensemble achieves 97.37% accuracy, 97.26% precision, 98.61% recall, 97.93% F1score, and 99.60% AUC-ROC — outperforming all individual base classifiers and prior ensemble work on this benchmark. Ten-fold stratified cross-validation confirms stability at 97.37% ± 2.39%. Robust Scaler preprocessing is introduced as a key novelty for handling clinical outliers. The system is deployed as a zero dependency, real-time Clinical Decision Support System (CDSS).

DOI: https://doi.org/10.5281/zenodo.20180176

Understanding Cognitive Load And Emotional Adaptability In Human–AI Collaborative Work: Evidence From An Experimental Study

Authors: Mrinmoy Roy, Shivanand Pawar

Abstract: As Artificial Intelligence (AI) increasingly integrates into professional environments, understanding its psychological impact on human collaborators becomes crucial. This study investigates the emerging paradigm of Human-AI Ensemblement—the synergistic interaction between human professionals and AI systems—in corporate decision-making contexts. Conducted in Bangalore, India’s technology hub, the study employs a true experimental design to evaluate cognitive load, emotional adaptability, trust in AI, and decision confidence across three task environments: human-only, semi-AI, and full-AI collaboration. Fifteen mid-level corporate tech professionals participated in a pilot experiment using validated psychological tools (NASA-TLX, PANAS, Trust in AI scales). Results revealed significant differences in cognitive load and decision confidence across groups, with the full-AI group demonstrating lower cognitive strain and higher confidence. Correlational analyses further indicated strong positive associations between trust in AI and both performance accuracy and decision assurance. These findings underscore the need for psychological preparedness alongside technical AI training. The study lays groundwork for scalable research and practical frameworks in workforce development, emphasizing emotionally intelligent, cognitively adaptive, and trust-calibrated human-AI teaming.

DOI: https://doi.org/10.5281/zenodo.20184586

Adaptive Learning Framework For Concept Drift Mitigation In Real-Time Data Streams

Authors: Kasu Shashanth Kumar, K.Bharath Kumar, Katiki Bhuvaneswar Sai, Mr.J.Manivannan M.E

Abstract: Machine learning models deployed in real-world environments often operate on continuous data streams where the underlying data distribution evolves over time. This phenomenon, known as concept drift, degrades predictive performance and requires continuous adaptation. Existing approaches primarily focus on drift detection while relying on frequent retraining, which is computationally expensive and impractical for real-time systems. To address this limitation, this paper proposes an adaptive learning framework designed to detect, characterize, and mitigate concept drift in real-time data streams. The proposed framework integrates a drift detection module, a drift memory mechanism to store historical drift patterns, and a selective incremental model update strategy based on drift severity. This approach reduces redundant retraining while maintaining model stability and adaptability. Experimental evaluation demonstrates improved adaptability and computational efficiency compared to traditional retraining-based methods. The proposed framework provides a scalable solution for maintaining machine learning performance in dynamic streaming environments.

Performance Metrics Design And Implementation For The Cultural Genealogy Protocol (CGP)

Authors: Dr. Bayomock Linwa André Claude, Mr. Traore Check Abdoul Kader

Abstract: The Cultural Genealogy Protocol (CGP), conceived and designed by Dr. Bayomock Linwa Andre Claude and published in IJSET (2026, Vol. 14, Issue 1), defines a comprehensive framework for African-centered genealogy management. A first implementation of the CGP platform was produced by Mr. Coulibaly Monpi Kapo Darrell at IUGB, establishing the foundational data management infrastructure. However, the performance metrics module explicitly specified in the CGP protocol as a core objective had not yet been implemented. This article presents the design, mathematical formalization, and prototype implementation of this metrics framework as a contribution intended for integration into the existing CGP platform. Eight statistically grounded metric groups are introduced, covering demographic analysis, matrimonial statistics, longevity indexes, name frequency analysis with phonetic similarity detection, and paternal branch elder identification. A dual-perspective analytics dashboard is described, and validation results obtained on a 22-member, four-generation dataset are reported. All metrics were computed in under 300 milliseconds, confirming the feasibility and correctness of the proposed approach.

DOI: https://doi.org/10.5281/zenodo.20186354

Event Management Website: A Web-Based System for Automated Event Registration and Booking

Authors: Nishant Kumar, Adarsh Sharma, Ms. Amita Pathania, Ms. Suman Chandila

Abstract: Traditional event management processes in academic and organizational settings suffer from inefficiencies including data redundancy, double-booking errors, and limited scalability. This paper presents a web-based Event Management Website (EMW) designed to automate and streamline event registration, booking, and administration. The system employs a three-tier client-server architecture using HTML5, CSS3, and JavaScript for the frontend presentation layer; Node.js and Express.js for backend application logic; and MySQL as the relational database. The system supports two distinct user roles—regular users and administrators—and implements secure session-based authentication with bcrypt password hashing. Comprehensive testing including unit, integration, system, and user acceptance testing was conducted, achieving a 100% pass rate across 15 test cases. The proposed system reduces manual workload, eliminates booking conflicts, and provides a scalable foundation for future enhancements including AI-based event recommendations and mobile application development.

DOI: https://doi.org/10.5281/zenodo.20196256

AI-Based Chest Disease Detection System

Authors: Prachi D. Vighne, Trupti A. Dalimbe, Dnyaneshwari M. Khedekar, Sudarshan J. Sikchi

Abstract: Chest diseases such as pneumonia, tuberculosis, and COVID-19 continue to be major global health concerns, leading to significant morbidity and mortality. Accurate and early diagnosis is essential for effective treatment; however, traditional methods rely heavily on expert radiologists to interpret chest X-ray images, which can be time-consuming and subject to variability. In many regions, the shortage of skilled professionals further limits timely diagnosis, highlighting the need for automated and reliable solutions. This work presents an Adaptive Explainable AI framework for chest disease detection that combines deep learning, interpretability, and confidence estimation. A Convolutional Neural Network (CNN) is utilized to automatically learn relevant features from chest X-ray images and classify them into multiple categories, including COVID-19, pneumonia, tuberculosis, and normal cases. To address the lack of transparency in deep learning models, explainable AI techniques such as Grad-CAM and saliency maps are incorporated. These methods provide visual insights by highlighting regions in the X-ray image that contribute most to the model’s prediction. In addition, the system estimates prediction confidence using probability-based measures, allowing users to assess the reliability of the output. A severity estimation module is also included, which analyzes the extent of affected regions in the image to categorize the condition into levels such as mild, moderate, or severe. This adds practical value for decision-making and prioritization. The proposed system is implemented as a web-based application, enabling users to upload chest X-ray images and receive real-time predictions along with visual explanations and severity assessment. Experimental observations indicate that the model achieves satisfactory performance while improving interpretability and user trust. Overall, the framework provides a balanced approach between accuracy, transparency, and usability, making it suitable as a supportive tool in medical diagnosis.

Edge AI Powered CNN Model for Rice Bacterial Blight Detection

Authors: Mr. P. Loganathan, Aakash K V, Abinaya R, Hema M, Jeebika S H

Abstract: Rice is a primary food source for more than half of the global population. However, rice crops are highly susceptible to diseases such as Bacterial Leaf Blight (BLB), caused by BLB is caused by Xanthomonas oryzae pv. oryzae, which can reduce yields by up to 70% under severe conditions. Traditional detection methods rely on manual inspection and laboratory testing, which are time-consuming, costly, and require expert knowledge unavailable to most rural farmers. This paper proposes an Edge AI-based Convolutional Neural Network (CNN) model for detecting rice bacterial blight in real time. The system allows farmers to upload rice leaf images through a web interface, where a trained CNN model analyzes and predicts the disease instantly. The model is optimized using lightweight architectures such as MobileNetV2 and EfficientNet, making it suitable for edge devices with limited computational resources. The proposed system achieves up to 97.2% classification accuracy with sub-second inference on mobile hardware, significantly outperforming traditional approaches. It improves detection speed, accessibility, and farmer engagement, enabling timely preventive actions and meaningful reduction in crop loss.

Feature Selection And Model Performance Analysis

Authors: Ms Dimpy, Vanshika Dubey, Sakshi Kumari, Aarzia Juned

Abstract: Feature selection is an important step in machine learning that helps improve model performance by removing irrelevant and redundant data. This study compares different feature selection techniques and analyzes their impact on various models. Results show that proper feature selection reduces complexity, improves accuracy, and prevents overfitting, leading to better overall model performance.

DOI: https://doi.org/10.5281/zenodo.20198033