Wellness Detection Using Clustered Federated Learning

Abstract

One of the most prevalent mental illnesses, depression, can present with various symptoms, making a clinical diagnosis and pathological research challenging. Even while technical documents show artificial intelligence can aid in treating and diagnosing depression, conventional centralised machine learning demands the collection of patient data, which restricts the therapeutic application of machine learning algorithms. We use federated learning to evaluate and diagnose depression to solve the concern of patient privacy regarding their medical history. To enable federated learning across institutions or parties, we first provide a cluster-based federated learning framework built on multi-source data that can be used to improve any standard machine learning model. The performance of the federated framework in comparison to other cooperative learning frameworks is examined and explained in the simulation section.