Application of machine learning models to predict malaria using malaria cases and environmental risk factors

Abstract

Malaria remains a significant cause of deaths and illness especially in sub-Saharan Africa (SSA). The efforts to eliminate malaria include the use of intermittent preventive prophylaxis (ITPp), indoor residual spraying (IRS), long-lasting insecticide-treated nets (LLINs), malaria prevention strategies and behavioural change education. Among these initiatives, predicting malaria cases at the ward level tremendously assist in malaria elimination, yet its application is still low. Therefore, this paper applied logistic regression, decision trees classifier, support vector machine, and random forest classifier to predict malaria in Buhera district. The study shows that logistic regression performs better, with 83% accuracy, 82% precision and 90% F1-score than other machine learning classifiers when predicting malaria outbreaks using environmental risk factors. These models can aid decision-makers to effectively allocate resources, development of malaria early warning systems, optimize the distribution of indoor residual spraying teams and spraying equipment, giving more priority to high sporadic areas.