A Federated Learning Model for Privacy-Preserving and Personalized Mental Health Monitoring using Big Data Analytics

Abstract

Mental health disorders are rising globally, posing challenges in early detection, personalized care, and data privacy. The conventional centralized Machine Learning (ML) models involve the transfer of sensitive patient information to central servers, a factor that jeopardizes privacy and reduces user confidence. In addition, these models normally do not consider individual differences in behavior and thus generalized predictions are made, which are less accurate. To address these shortcomings, this research proposed a Federated Learning (FL)-based model of mental health monitoring that allows training models without raw data sharing in distributed medical customers. The model incorporates the concept of Big Data Analytics (BDA) to handle heterogeneous mental health records and assist with user-level customization and meaningful evaluation based on overall performance measures. The simulation outcomes indicate that the proposed model has a global accuracy of 99.53% and a miss rate of just 0.47% which beats conventional centralized methods. This research highlights a scalable, secure, and personalized solution for mental healthcare, advancing the integration of federated learning and big data in digital health systems.