FEDERATED LEARNING FRAMEWORK FOR PRIVACY-PRESERVING HEALTHCARE ANALYTICS

• Dr. Amelia Richardson

  Department of Computer Science, University of Helsinki, Finland

  Author

Healthcare institutions collect massive volumes of sensitive patient information that hold immense potential for predictive analytics. However, traditional machine-learning techniques require centralized data pooling, which exposes institutions to risks of privacy breaches, data misuse, and regulatory non-compliance. Federated Learning (FL) provides a decentralized alternative that enables model training across distributed datasets without transferring raw patient records. This paper presents an enhanced federated learning framework designed specifically for healthcare analytics, offering improved privacy, accuracy, communication efficiency, and robustness against poisoning attacks. Using real hospital datasets (simulated for this study), the proposed framework achieves a 9–14% performance improvement compared to baseline FL methods. Results demonstrate that the model preserves patient privacy while enabling effective diagnostic predictions. The study concludes that federated learning is a promising and practical architecture for large-scale, privacy-preserving healthcare systems.

• pdf

This work is licensed under a Creative Commons Attribution 4.0 International License.