Prediction of malaria incidence using climate variability and machine learning

Malaria remains a serious obstacle to socio-economic development in Africa. It was estimated that about 90% of the deaths occurred in Africa, where various factors such as ecosystem and climate conditions are favorable to species of mosquitoes transmitting the malaria parasite. Some malaria epidemic prediction systems have been built to mitigate the increase of the disease outbreaks in some African countries; however, there is a need for better models with improved prediction ability based on non-seasonal variations in climatic conditions. This research proposes a machine learning-based model for the classification of malaria incidence using climate variability across six countries of Sub-Saharan Africa over a period of twenty-eight years. The work begins with a feature engineering process, which identifies the climate factors that affect the incidence of malaria, followed by the k-means clustering process for outlier detection, and then, XGBoost algorithm for classification. The results suggest that although the exact association between malaria incidence and climate variability varies from one geographic region to another, the non-seasonal changes in three climatic factors (precipitation, temperature, and surface radiation) significantly contribute to the outbreak of malaria. The proposed system was compared with other classification models, and the comparative results showed that the proposed system outperformed other classification models. The malaria incidence classification model is an early detection mechanism that helps to monitor the spread of malaria; it is a unique data-driven knowledge discovery system that will assist public health authorities in learning the effects of climate factors on health and also in developing relevant preventive and adaptive mechanisms to ensure a timelier health service in order to save lives.