Machine Learning Algorithms for the Prediction of the Spread of COVID-19 in Namibia

Abstract

Improving the accuracy and stability of daily COVID-19 forecasts is crucial for effectively managing and controlling the pandemic. This chapter compares the performance of different machine-learning algorithms in predicting and forecasting the spread of COVID-19 in Namibia. Machine learning approaches that include the support vector machine (SVM), the TBATS, the generalized additive model (GAM), and the Stochastic Gradient Boosting Machine (SGBM) approach are compared. Selection of the best-performing model is done using plots of forecasts from fitted models on the test dataset since plots are visually appealing. A further selection of the best model is done using key performance indicators (KPIs), that is, root mean square error (RMSE), mean absolute percentage error (MAPE), and the coefficient of determination R2. Results show that the positive rate, reproductive rate, and stringency index contribute significantly (p-values<0.05) to the spread of COVID-19 in Namibia. From the fitted models GAM and the SVM linear kernel function are the best performers in forecasting daily COVID-19, although based on KPIs GAM outperforms the SVM linear kernel function. This study recommends the use of both models to help solve the forecasting problem and the identification of significant regressors. Accurate prediction and forecasting help in giving the health sector early warning signs and preparedness to help manage and control epidemics. This will go a long way in helping achievement of the Sustainable Development Goal number 3 of health and wellness.