International Journal of Science, Engineering and TechnologyAuthors: A. Verma, A. Rastogi, A. Saini, M. Saini
Abstract: As machine learning (ML) models are increasingly deployed in mission-critical domains such as healthcare, finance, and cybersecurity, the opacity of complex, “black-box” models raises concerns about trust, regulatory compliance, and safety. Explainable AI (XAI) seeks to bridge this gap, yet scalable, user-friendly, and cloud-integrated solutions for model interpretability remain scarce. This paper presents the SHAP-Driven Feature Attribution Analyzer—a comprehensive, cloud-native XAI framework that leverages SHAP (SHapley Additive exPlanations) for transparent and actionable feature attribution. Our motivation is to democratize high-fidelity explainability, making ML model decisions interpretable to diverse stakeholders while ensuring seamless integration with modern cloud platforms. We describe an end-to-end system architecture combining Next.js and Recharts for interactive visualization, a Python-based SHAP computation engine, and Firebase for secure, scalable backend services. The workflow spans data ingestion, preprocessing, model training (covering logistic regression, random forest, and XGBoost), SHAP-based explanation, and intuitive visualization pipelines. We deeply examine SHAP’s theoretical underpinnings—rooted in cooperative game theory—and its practical computation (TreeSHAP, KernelSHAP), addressing both global and local interpretability and the handling of feature collinearity. Our experiments demonstrate the efficacy of the Analyzer using benchmark datasets, with results visualized via summary, beeswarm, force, and waterfall plots. Key contributions include a modular, cloud-integrated architecture, robust security design, and extensive empirical validation. By lowering the barriers to XAI adoption and deployment, this work addresses critical gaps in accessibility, interpretability, and operationalization of explainable ML, paving the way for transparent, fair, and trustworthy AI systems.
