Hybrid Federated Learning for Secure and Accurate Heart Disease Prediction

Cardiovascular disease continues to be a leading cause of death globally, emphasizing the need for advanced predictive models to facilitate early detection. Traditional ma chine learning techniques frequently depend on centralized data aggregation, which presents notable challenges regarding privacy and data security. Federated Learning (FL) mitigates these concerns by facilitating decentralized training across multiple clients, ensuring data privacy is maintained. This study presents a Hybrid Federated Learning (FL) model that combines deep learning with the Federated Averaging (FedAvg) algorithm to effectively aggregate updates from local models. The proposed model is evaluated using the Framingham Heart Study dataset, with its performance compared against that of a centralized deep learning model. The findings reveal that the Hybrid FL model achieves comparable accuracy while ensuring data privacy, underscoring its potential for practical medical applications.