Web Based Multi Feature Propagation Analysis Model for Efficient Disease Interference and Recommendation Using Deep Learning

Abstract

Accurate prediction of chronic diseases is critical for proactive healthcare management. This paper proposes a Multi-feature Propagation Analysis-based Deep Learning Model (MPADM) to address this challenge. The model integrates diverse patient data, including medical, diagnostic, genetic, and historical features, collected from multiple sources. After preprocessing, the network is trained to calculate distinct Propagation Weights (PWs) for each feature category, Diagnosis Propagation Weight (DPW), Genetic Propagation Weight (GPW), and Historical Propagation Weight (HPW). These weights, estimated across different disease classes, are aggregated to generate a final predictive score for chronic diseases. To support clinical decision-making, the model also computes a Treatment Support (TS) metric, ranking hospitals and medical practitioners for user recommendation. Implemented with a web-based interface for accessibility, the MPADM model demonstrates enhanced efficacy, significantly improving prediction accuracy and the quality of therapeutic recommendations compared to existing benchmarks.