Evaluation of Supervised machine learning classifiers for detecting the degrees of possibility of Coronavirus disease infection

Abstract

The COVID-19 virus spread from China throughout the world, causing out of control challenges to the public health community. During the epidemic, it is difficult to act against infectious disease due to its unknown trends, so the spread prediction becomes difficult in light of the scarce data. With the absence of treatment for COVID-19 infection, countries have taken some mitigation steps and policies, such as general lockdown and social distance measures, but there has been variation in the extent of viral spread due to several additional factors. Prediction methods based on Artificial intelligence (AI) and Machine Learning (ML) can help in suggesting new policies or even assessing the effectiveness of the existing ones. Such methods have attracted wide attention from researchers implementing statistical modeling and machine learning methods. The objective of this study is to examine different supervised classification approaches to detect the degrees of possibility of Coronavirus disease infection in different countries. Naïve Bayesian Classifier (NBC), Decision Tree Classifier (DTC), Linear Discriminant Analysis (LDA), Support Vector Machine (SVM) and Artificial Neural Network classification (ANN) are machine learning algorithms used for the prediction of coronavirus disease cases according to the time of their evolution while considering data collected from official reports and scientific journals. Since we collected mixed data, we suggested to also apply the supervised classifiers available for mixed data, such as Extended Gamma (EG) and Naive Associative Classifier (NAC). The results showed that ANN and DTC supervised algorithms allow better discrimination between the degrees of possibility of Coronavirus disease infection among advanced methods like NBC, SVM, NAC, EG and LDA.