An Investigation of the Coronavirus Disease (COVID-19) Mortality Risk Using Machine Learning

Abstract

Septic shock, acute respiratory distress syndrome, multi-organ failure are all possible complications of the illness. In this work, machine learning is used to construct and assess mortality risk models that were evaluated (both positive and negative). The authors employed machine learning to gather information about 51,831 individuals (of which 4,769 were confirmed cases of COVID-19). The data collection comprises data from the next week (47,401 tested and 3,624 confirmed COVID-19). It is still uncertain if the SVM classifier ensemble can beat single SVM classifiers regarding the number of positive and negative predictions made in COVID-19. This investigation will investigate the accuracy of ensembled SVM and simple SVM on small and large COVID-19 datasets. The ROC, accuracy, F-measure, classification, and calculation time of SVM and SVM ensembles are evaluated and compared. According to the data, linear-based SVM performs the best when used as a bagging strategy. When dealing with tiny datasets that need feature extraction during pre-processing, bagging and boosting SVM ensembles may benefit.