A Classification Framework to Detect Sars Covid-19 Disease Using Feature Selection and Variant-Based Ensemble Learning

Abstract

The hazardous COVID-19 pandemic has caused millions of deaths worldwide which depicts the significance of an early screening of this infection in order to stop it from spreading. Real-time polymerase chain response (RT-PCR) test has been in use for detection of COVID-19 infection but it is time consuming and might generate results rather late as the test samples are required to be examined and tested at a suitably equipped lab. In contrast to this methodology, many deep learning strategies have been put forward to successfully identify this sickness by inspecting CT scan images and chest X-Ray images but doses of radiation, high cost and lack of specialized equipment do not allow these methods to be generally used for the prediction of COVID-19. In this research a classification framework is proposed which is inexpensive, readily available, and considerably fast and does not require specialized equipment or laboratories as it is based on complete blood count results. This proposed model uses machine learning techniques and routine blood results to predict COVID-19. It comprises of three stages 1) data preprocessing 2) hybrid feature selection (FS) 3) Ensemble learning classification.