Predicting malaria incidence using case load and metrological data in Sidama Regional State

Abstract

Malaria remains a significant public health challenge, particularly in tropical regions like Ethiopia's Sidama Regional State, where climatic factors heavily influence transmission dynamics. This study utilizes an ANN feed forward model to integrate meteorological data (minimum temperature, maximum temperature, and rainfall) with historical malaria case records (2017–2022) to construct a predictive model for malaria incidence. Data were obtained from the Ethiopian National Meteorological Agency and the Sidama Regional Health Bureau. Four districts Boricha, Dale, Hawassa Zuria, and Shebedino were used to validate the model. To determine the most effective machine learning technique for malaria prediction, this study compared the ANN feed forward model with Random Forest and Decision Tree models. Among these, the ANN feed forward model demonstrated superior predictive accuracy, achieving the lowest RMSE values across districts, with Shebedino (0.4787) and Hawassa Zuria (0.7359) performing best. However, challenges remain in capturing short-term fluctuations, particularly in Boricha (RMSE: 2.610).The results emphasize the importance of incorporating meteorological factors into malaria prediction models and highlight the ANN model's potential as a robust early warning system. By enabling public health officials to forecast outbreaks and allocate resources more effectively, predictive models like ANN can significantly enhance malaria prevention efforts. Future research should focus on improving model accuracy by integrating additional variables and exploring advanced machine learning techniques to handle complex transmission scenarios