Regression Models in Malaria Cases Prediction Using Climatic Data

Abstract

Malaria disease remains a challenging disease to eradicate and still a danger to human population globally. A 2022 report by World Health Organization, indicated the number of malaria cases registered in 2020 were 241 million in comparison to 2019 where 227 million. Different studies have been conducted globally and show factors that cause malaria infections in numerous regions of the globe are the same. Climatic conditions factors such as temperature, humidity and rainfall account for these cases. There is still a challenge by health related institutions is combating the malaria disease for lack of tools that can be used to predict the disease occurrences. With a tool that can help them to do prediction, they will be in a position to put in place the right preventive measures proactively. The intention of this investigation is to develop and recommend a model based on artificial intelligence. It can be used for malaria infection cases prediction, based on historical data for malaria infection cases and climatic conditions variables. Three years data (2019 to 2021) for malaria infections and climatic data (which includes temperature and rainfall) was sourced. The historical data was used to fit most preferred machine learning algorithms, namely, Random Forest Regressor (RFR), Support Vector Regressor (SVR), and Extreme Gradient Boosting (XGB) regressor algorithms to determine the best prediction algorithm. Best performing algorithm was recommended based on performance evaluation done using the recommended metrics, namely, R-squared (coefficient of determination), Root Mean Squared Error, Mean Squared Error and Mean Absolute Error. Among the three regressor models, SVR was observed to have better performance compared to RFR and XGB with MAE of 44.98, MSE of 2594.67, RMSE of 50.94, and R-squared of 0.895. According to this study, it is recommended to use SVR for malaria infections cases prediction where there is low volume of dataset available. By extrapolating the same historical data, it was proven both RFR and XGB perform better than SVR on big dataset. This research has demonstrated that using machine learning algorithm models for predicting malaria cases would contribute in providing insightful information for health related organizations and institutions to assist them in planning, preparing and setting up the right interventions proactively to contain or prevent adverse effects of malaria disease on the population