Comparing machine learning methods for predicting COVID-۱۹ mortality and ICUs Transfer

The current study's main goal was to compare the performance of four machine learning (ML) methods for predicting COVID-۱۹mortality and ICUs transfer. In this retrospective cohort study, information of ۳۲۹ COVID-۱۹ patients was analyzed. These patientswere hospitalized in Besat hospital in Hamadan province, the west of Iran. The support vector machine (SVM), least-square SVM(LS-SVM), random forest (RF), and Naïve Bayes (NB) were used for predicting COVID-۱۹ mortality and ICUs transfer. Thesemethods' performance was evaluated using sensitivity, specificity, positive predictive value, negative predictive value, andaccuracy. Of the ۳۲۹ COVID-۱۹ patients, ۵۷ (۱۵.۵%) died, and ۱۰۶ (۳۲.۲%) were transferred to ICUs. Among ML methods, RFperformed the best with the highest accuracy (۰.۸۵) for predicting COVID-۱۹ mortality. Moreover, SVM had the highest accuracyfor predicting ICUs transfer. This study showed that the performance of RF and SVM provided better results compared to othermethods for predicting COVID-۱۹ mortality and ICUs transfer, respectively.

Keywords:

COVID-۱۹ , Mortality , Intensive Care Units , Machine Learning.