INTELLIGENT DATA MINING FRAMEWORK FOR HEALTHCARE PREDICTIVE ANALYTICS USING REAL-WORLD EVIDENCE: ENHANCING CLINICAL DECISION-MAKING THROUGH SCALABLE AND EXPLAINABLE MODELS

Abstract

The convergence of Big Data, Artificial Intelligence, and medical informatics has created a transformative opportunity for Healthcare Predictive Analytics. However, the clinical integration of high-performance models is currently stalled by the "transparency-performance" trade-off and the inherent difficulty of scaling complex algorithms to process heterogeneous Real-World Evidence. This research presents the Intelligent Data Mining Framework for Healthcare, a scalable and explainable end-to-end architecture designed for enterprise clinical deployment. By leveraging state-of-the-art interoperability standards like Fast Healthcare Interoperability Resources and advanced feature engineering techniques, the framework manages the high-dimensional complexity of large-scale databases including MIMIC-IV and the CHoRUS Bridge2AI dataset. We introduce a hybrid feature selection mechanism combining Chi-Square and Principal Component Analysis to optimize predictive performance. Experimental validation across multiple cardiovascular and metabolic disease benchmarks demonstrates a superior predictive accuracy of up to 98.7% while providing robust, clinician-interpretable explanations via SHAP and LIME. The findings suggest that the IDM-HPA framework significantly enhances clinical trust and provides a standardized foundation for the next generation of digital health systems.