Decentralized machine learning for disease outbreak prediction: Enhancing data privacy with federated learning

The ability to predict and contain disease outbreaks is essential for global public health. However, traditional machine learning models for epidemiological forecasting relieve centralized data aggregation, which poses significant privacy risks and regulatory challenges. In this study, we propose a federated learning (FL)-based decentralized framework that enables collaborative model training across multiple healthcare institutions without exposing sensitive patient data. By leveraging privacy-preserving techniques such as secure aggregation and differential privacy, our approach ensures data confidentiality while maintaining predictive accuracy. We evaluate our framework using real-world datasets from multiple healthcare agencies and demonstrate that it achieves performance comparable to centralized models while significantly reducing privacy risks. Our findings highlight the potential of federated learning to enhance cross-institutional collaboration in public health while addressing critical privacy and security concerns. This work underscores the importance of decentralized AI-driven solutions for epidemiological forecasting and privacy-preserving healthcare analytics.