Dengue fever has become an increasingly significant public health concern in Nepal, with outbreaks showing strong seasonal patterns influenced by climatic conditions. Temperature, rainfall, and relative humidity affect mosquito breeding dynamics and virus transmission, making climate variables important predictors of dengue incidence. While previous studies in Nepal have primarily examined statistical relationships between climate variability and dengue cases, predictive modeling approaches remain limited. This study develops and compares climate-based machine learning models to predict dengue incidence in Nepal.
Monthly dengue case data and climatic variables, including temperature, rainfall, and relative humidity, were collected for the period 2022–2025. To capture delayed climatic effects on vector ecology and disease transmission, one-month and two-month lagged climate variables were incorporated into the modeling framework. Comparative prediction models were developed using ensemble machine learning techniques, including Random Forest and Gradient Boosting, to model nonlinear relationships between climate variables and dengue incidence.
Model performance was evaluated using Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) to assess predictive accuracy. The results demonstrate that machine learning models can effectively capture the complex relationships between climatic factors and dengue incidence. Incorporating lagged climate predictors improved forecasting performance, highlighting the importance of delayed climatic influences on dengue transmission dynamics.
The findings demonstrate the potential of climate-based machine learning approaches for dengue prediction in Nepal and highlight their usefulness for developing early warning systems to support public health preparedness and outbreak prevention.