Privacy-Preserving Federated Learning for Heart Disease Prediction under Non-IID Healthcare Data

This research paper presents a privacy-preserving Federated Learning approach for heart disease prediction under Non-IID healthcare data conditions. The proposed approach simulates multiple healthcare clients (hospitals) that collaboratively train a global machine learning model without sharing raw patient data. The study focuses on Federated Averaging (FedAvg) for model aggregation and Differential Privacy to improve patient data confidentiality. It also considers Explainable AI to support the interpretability of model predictions. The work evaluates the model using performance metrics such as accuracy, precision, recall, F1-score, and AUC. Client-wise performance and the trade-off between privacy and prediction accuracy are also considered. The study aims to demonstrate how Federated Learning can support collaborative healthcare prediction while preserving data privacy across distributed healthcare environments.