Predicting the Existence of COVID-19 using Machine Learning Based on Laboratory Findings

Abstract

Since December 2019, a new coronavirus disease (COVID-19) was detected in Wuhan, China, spread all over the world. Many research papers have been published to study this disease and help humans to overcome this pandemic. Here, we highlighted the prediction process of COVID-19 based on a combination between wrapper feature selected (FS) algorithm and four different classifiers, namely Convolutional Neural Network (CNN), decision trees (C4.5), nearest neighbors (kNN) and, Naïve Bayes (NB). A real dataset has been used in this paper generated by Hospital Israelita Albert Einstein at Sao Paulo, Brazil. The obtained results show an excellent performance of BGA with CNN compared to other methods with accuracy 76%.