Data for "Red Alert: Advancing Dengue Prevention with Environmentally and Socio-Demographically Informed AI Models"

Dengue, an arboviral disease transmitted by Aedes aegypti, remains a major public health challenge. In 2024, Brazil reported over six million cases, highlighting the urgent need for innovative methodologies to mitigate this scenario. Our objective was to forecast dengue incidence in Sergipe State and to investigate the contribution of environmental, socioeconomic, and climatic factors through Shapley Additive Explanations (SHAP). Dynamic ecological niche models of A. aegypti were generated with MaxEnt (2016–2022) to derive monthly climatic suitability, which was integrated with SINAN epidemiological data, TerraClimate variables, MapBiomas land-use data, and socioeconomic indicators from IBGE. Eight predictors were selected (climatic suitability, minimum temperature, precipitation, surface runoff, forest cover, pasture area, Gross Domestic Product—GDP, and sanitation). Four ML algorithms (eXtreme Gradient Boosting—XGBoost, Random Forest, Support Vector Machine, and Deep Learning) were trained (80% of data, 100 iterations, 5-fold CV) and tested (20% of data). XGBoost achieved the best performance, with 73.4% precision in predicting high-incidence categories and metrics above 0.5 across evaluations (accuracy, precision, recall, and F1-score), demonstrating its potential as a reliable tool for dengue monitoring. SHAP analysis revealed GDP, forest cover, and precipitation as the most influential predictors. Low GDP values were positively associated with dengue incidence, reflecting inadequate urban infrastructure. Reduced forest cover contributed to increased vector habitats, while higher precipitation favored breeding site availability and vector–host interactions. These findings underscore the value of ML approaches for arboviral disease surveillance, offering a scalable framework for continuous spatial monitoring and decision support in public health.