DYNAMIC ENSEMBLE MODELS FOR CLIMATE-DRIVEN EPIDEMIC PREDICTION

Epidemics driven by climate variability pose critical challenges to public health systems, as factors such as rising temperatures, disrupted seasonal cycles, and extreme weather events substantially impact disease transmission dynamics. This study introduces a novel framework that combines deep learning with graph-based modeling to predict Respiratory Syncytial Virus (RSV) incidence. By integrating climate, epidemiological, and socioeconomic data across diverse U.S. states, our approach tackles the challenges of generalizing models across geographically varied regions. Leveraging a dynamic ensemble technique enhanced by transfer learning, the framework incorporates temporal, spatial, and climatic dependencies to achieve robust, region-specific time-series forecasts. The model not only captures localized trends but also adapts to heterogeneous regional patterns, offering a scalable and accurate solution for epidemic forecasting. These findings highlight the potential of data-driven methodologies to advance climate-resilient public health strategies, enabling real-time monitoring, targeted interventions, and resource optimization.