A Novel Approach: Hybrid Feature Selection Using Updated Gini Index with MapReduce and Federated Learning for Chronic Disease Management

Chronic diseases pose significant healthcare challenges, necessitating efficient predictive analytics powered by Machine Learning (ML) and Internet of Things (IoT) technologies. However, handling high-dimensional healthcare data presents scalability and privacy concerns. This paper proposes a novel hybrid feature selection framework that integrates MapReduce for scalable computation and Federated Learning (FL) for privacy-preserving model training. The Updated Gini Index is introduced to enhance feature selection accuracy while reducing computational overhead. The proposed model is evaluated using real-world IoT healthcare datasets, demonstrating a 35% reduction in computational cost and a 15.7% improvement in accuracy over traditional methods. Results validate the efficacy of our approach in enabling real-time, secure chronic disease prediction. Future work will explore optimization techniques to further enhance model performance and deployment scalability.