Prediction of Covid-19 confirmed cases and deaths using hybrid support vector machine-Taguchi method

Abstract

With the emergence of the COVID-19 pandemic, experts aim to predict the number of confirmed cases and take measures to reduce death rates in order to be prepared for future pandemics. Determining the variables that affect the confirmed cases and deaths and predicting the number of confirmed cases and deaths is important for taking precautions quickly.Many variables are affecting the confirmed cases and death rate of COVID-19 and causing the uncertainty in the process to increase even more. In this study, 19 variables affecting the number of confirmed cases and deaths from COVID-19 were determined. These variables are used as input in the support vector machine (SVM) to predict the number of confirmed cases and deaths. The hyperparameters of SVM, such as kernel type, C, and ɣ values, were determined by the Taguchi method. The hybrid method determines the optimum hyperparameter values of SVM in a short time and is less labor-intensive compared with other hyperparameter tuning methods.The application dataset was obtained from 51 countries' data. At the end of the study, the hybrid SVM-Taguchi method determined the number of confirmed cases and deaths by 91.8% and 94.2%, respectively. Besides, the hybrid SVM-Taguchi method was compared with different artificial intelligence methods. The hybrid SVM-Taguchi method gave better results according to the statistical performance criteria.