Africa's Malaria Epidemic Predictor: Application of Machine Learning on Malaria Incidence and Climate Data

The 2019 World Malaria Report confirms that Africa continue to bear the burden of malaria morbidity. The continent accounted for over 93% of the global malaria incidence reported in 2018. Despite the numerous multi-level and consultative efforts to combat this epidemic, malaria continues to claim thousands of human lives, especially those of children under 5 years of age. Since malaria is preventable and treatable, one of the solutions towards reducing the number of deaths is by implementing an effective malaria outbreak early warning system that can forecast malaria incidence long before they occur. This way, policymakers can put mitigation measures in place. Tapping into the success of machine learning algorithms in predicting disease outbreaks, we present a malaria outbreak prediction system that is anchored on the well-established correlation between certain climatic conditions and breeding environment of the malaria carrying vector (mosquito). Historical datasets on climate and malaria incidence are used to train nine machine learning algorithms and four best performing ones identified based on classification accuracy and computation performance. Preceding the models' development, reliability and correlation analysis was carried out on the data; this was then followed by reduction of the dimensionality of the feature space of the two datasets. Given the power of deep learning in handling selectivity variance, the malaria predictor system was developed based on the deep learning algorithm. Further, the evaluation of the system was done using the Simulator function in RapidMiner and the accuracy of the predictions assessed using an independent dataset that was not used in the models' development. With prediction accuracy of up to 99%, this system has the potential in contributing to the fight against malaria epidemic in Africa and elsewhere in the world.