Prediction of COVID-19 using Machine Learning Models based on Clinical Blood Test Data

Abstract

The global pandemic of Coronavirus Disease 2019 (COVID-19) has caused serious problems and threatened the lives of many people. To effectively combat the disease, early and precise screening of infected individuals is essential. The study uses blood test data which comprises 1736 instances and 35 features that have been collected from the patients who were admitted to the emergency department at the San Raffaele Hospital. For predicting COVID-19 in patients, RT-PCR tests a-re widely used. Once a patient has been identified with the presence of COVID-19, the patient should approach a healthcare professional to determine the severity of the virus and appropriate medical treatment and supportive care should be provided. The patient’s condition should be closely monitored to ensure that their health is improving and to detect any complications that may arise. For this purpose, blood test samples taken from the patient will help to diagnose his condition and the severity of the virus. In this work, a feature selection technique known as Recursive Feature Elimination (RFE) has been used to find out the optimal set of features that are highly related to the existence of COVID-19 in patients. The features obtained using RFE are then applied with a machine learning model and the best results are achieved using a Random Forest classifier with an accuracy of 89%.