Privacy-Focused Federated Learning for Mental Health Data with XAI Techniques
V. Meena, G. Raghavender, Shriram Jayakar, J. Senthil Kumar · 2025
Protecting the privacy of sensitive data is paramount in the field of mental health care, keeping in mind the amount of data that the healthcare applications produce. In such cases, federated learning comes into the picture, which has emerged as a leading solution for training decentralized models while preserving the confidential information from the models. To overcome some vital issues, we have come up with an approach of solving the problem using Federated Learning which is communication-efficient and specifically designed for mental healthcare applications. This model's framework depends on federated averaging that involves averaging of weights to update the models' parameters. We evaluate the model's working on a classic mental health care dataset called WESAD from the UCI ML repository with identically distributed data. Our results highlight the promise of the FL model, demonstrating its potential to transform the field. To gain confidence in our results, we have utilized explainable artificial intelligence, for tabular data so that we can be sure of our model's predictions. The model performed well with an overall accuracy of 92.56%, a precision score of 92.67%, a recall of 92.56%, and an F1 score of 92.25%.