Fine-scale predictive modeling of Aedes mosquito abundance and dengue risk indicators using machine learning algorithms with microclimatic variables

Effective prediction of Aedes mosquito abundance and dengue risk indicators such as the Aedes Index (AI) and Dengue Positive Trap Index (DPTI) is essential for early intervention and targeted vector control. However, current models often rely on coarse regional data and fail to account for microclimatic variations, limiting their predictive accuracy in dengue hotspots. This study developed fine-scale predictive models using machine learning algorithms; Artificial Neural Networks (ANN), Random Forest (RF), and Support Vector Machines (SVM) to estimate mosquito abundance and dengue risk at the species level based on daily microclimatic data (temperature, relative humidity, and rainfall) collected over 26 weeks in Kuala Selangor, Malaysia. Predictor variables included single, dual, and triple combinations of microclimatic inputs, and models were trained and validated using 10-fold cross-validation and a 70:30 train-test data split. ANN consistently outperformed RF and SVM in predicting the Aedes Index (e.g., Ae. aegypti: MAE = 0.175, RMSE = 0.248), while RF and SVM demonstrated superior performance in DPTI predictions for Ae. aegypti and Ae. albopictus, respectively. Models incorporating dual or triple microclimatic variables yielded significantly lower error metrics than those using single predictors. Rainfall emerged as the most influential single factor across species. Variations in model performance were likely due to species-specific responses to environmental conditions and the nonlinear interactions captured by the algorithms. Compared to benchmarks in related tropical settings, the reported error metrics demonstrate improved prediction accuracy. Integrating time-lagged microclimatic variables into machine learning frameworks enhances the predictive accuracy of dengue risk indicators at a fine spatial scale. These models provide a data-driven basis for proactive dengue control strategies, enabling timely interventions tailored to specific mosquito species and environmental triggers.