Enhancing Malaria Incidence Prediction Using Weather Variables: A Comparative Study of Gradient Boosting and Random Forest Regression Models

Malaria remains a significant global health challenge, particularly in regions where environmental conditions favor the Anopheles mosquito vector. This study evaluates the effectiveness of incorporating weather variables into predictive models for malaria incidence, comparing Gradient Boosting Regression (GBR) and Random Forest Regression (RFR) techniques. Researchers utilized daily weather records (temperature, humidity, and rainfall) and corresponding malaria incidence data from ten districts in Rwanda, spanning April 2019 to June 2023. Both GBR and RFR models demonstrated strong predictive performance when weather variables were included. GBR slightly outperformed RFR, achieving a Root Mean Squared Error (RMSE) of 671.52 and an R-squared value of 98.20%, compared to RFR's RMSE of 740.47 and R- squared of 97.81%. These results represent a significant improvement over models without weather variables, with increases in R-squared values from 93.58% to 98.20% for GBR and from 93.28% to 97.81% for RFR. Our findings underscore the crucial role of meteorological data in enhancing malaria prediction accuracy and highlight the potential of machine learning algorithms in capturing complex, non-linear relationships between environmental factors and malaria transmission dynamics. This research contributes to ongoing efforts to improve malaria surveillance and control strategies, offering a promising approach for more accurate and timely predictions of disease risk in endemic regions.