Feature data and missingness for the regions.csv.

BackgroundDengue is a significant global health threat, transmitted by mosquitoes and influenced by multiple factors. A comprehensive analysis of the impact of these factors on dengue at a global scale is helpful for better understanding and effective control of dengue epidemics.MethodsThis study employed machine learning techniques to develop a global predictive model for forecasting annual dengue cases. A wide range of multi-source features, including historical cases, population, climate, air travel, forest, anemia, vector, serotype and socioeconomic features, were comprehensively considered. The impact of these features was revealed using the SHAP (Shapley Additive Explanations) framework.ResultsThe global multi-variable model outperformed the baseline model, indicating the importance of considering multiple factors. Among the multi-source features, historical cases contribute the most, at about 73.63%. Risk factors associated to dengue were identified, including the occurrence of Aedes mosquitoes, changes in the predominant serotype, and the prevalence of anemia. Feature contribution pattern was different between hyperendemic and non-hyperendemic regions. In hyperendemic regions, historical cases and population were found to contribute more significantly, emphasizing the role of population immunity in dengue dynamics.ConclusionsDengue is influenced by a wide range of multi-source factors, and prevention and control measures should be specifically designed while taking into account regional differences for effective control of dengue.