Comparing Different Machine Learning Algorithms for Predicting Coronavirus (Covid-19) Disease

Abstract

COVID-19 is a viral and pandemic disease faced the whole world between the end of year 2019 and the beginning of year 2022. firstly, appeared in China and then spread out all over the world and became a global threat to human health patients that had COVID-19 caused serious symptoms thus several of them die due to a part of their organ failure especially liver. This paper used algorithms of the machine learning to construct the COVID-19 severe ness apprehension model. Four machine learning classification techniques were evaluated: Logistic Regression (LR), Support Vector Machine (SVM), Random Forest (RF) and Extreme Gradient Boosting (GXB). In aiming to treat the imbalance classification, Synthetic Minority Oversampling Technique (SMOTE) was utilized. Two set of investigation have been built with original dataset and with the SMOTE sampling technique. Based on several metrics for evaluation, Random Forest and Support Vector Machine Classifier has shown the highest performance for both datasets without and with SMOTE while the minimum result achieved by Logistic regression also for both datasets. Furthermore, the achievement performance of the four machine learning models experienced with SMOTE is strongly preferable than performance of classifiers competent without SMOTE. Furthermore, Top 25, 20 and 15 features importance was conducted using ExtraTree-classifiers, the variation between the accuracy for different features selection were very small.