A New Two-Way Ensemble Prediction for Infectious Diseases

Abstract

Infectious diseases pose a persistent global health threat, making accurate forecasting essential for effective response. A deep understanding of their underlying mechanisms and the role of environmental factors is crucial to improving predictive models. In this study, we first investigate the relationship between key meteorological variables, including average temperature, relative humidity, PM2.5, and PM10 and the progression of infectious diseases, using COVID-19 data from Beijing and Shanghai as a representative case. A generalized additive model (GAM) is employed to determine the driven meteorological factors that affected the development of infectious diseases in general. Based on this analysis, we incorporate the identified meteorological variables into a modified susceptible–exposed–infected–recovered (SEIR) model. Then, further integrated this model into a novel two-way ensemble prediction (TEP) framework. This integrated method captures temporal and spatial dynamics of disease transmission. The proposed method demonstrated a significant improvement in prediction accuracy compared to traditional epidemiological models. This two-way integrated modeling framework offers a practical and rapid response to epidemic forecasting, providing valuable insights for future pandemic predictions, especially in regions with limited data on human mobility.