Real-Time Epidemic Forecasting Across Heterogeneous Populations: A Multi-Source Volatility-Gated Ensemble Framework
Abstract
Accurate epidemic forecasting at a global scale is inherently challenging due to heterogeneous reporting quality, behavioural variability, and differing rates of policy changes across countries. This study presents a comprehensive forecasting pipeline trained on multi-source COVID-19 data spanning among 253 countries from 2020 to 2024. Encompasses data preprocessing, feature engineering, and the development of both global and regional predictive models. The proposed Volatility-Gated Meta-Ensemble (VGME), which integrates XGBoost, Random Forest, and Ridge regression through a country-specific weighting scheme. A volatility-controlled integration serve as the final corrective mechanism to enable temporal dynamics. VGME achieves Root Mean Square Error (RMSE) reductions of more over 1,060 than stationary ensembles and 709 compared to volatility. Across all 253 countries, the model attains a mean correlation of 0.614 and a 50th-percentile accuracy of 0.889. Error and inter-metric analyses confirm strong magnitude tracking with minimal systematic bias, even under substantial cross-country variability.