Feature Extraction With Forest Classifer To Predicate Covid 19 Based On Thorax X-Ray Results

Abstract

Coronavirus 19 (COVID-19) is a highly contagious infection caused by the acute respiratory syndrome coronavirus 2 (SARS-CoV-2). COVID-19 is a new virus for which no cure has been found, marked by the increasing death rate worldwide. Coronavirus disease which can cause pneumonia which attacks the air sacs of the lungs with symptoms of dry cough, sore throat to acute respiratory distress (ARDS) that occurs in COVID-19 patients. One of the ways to detect the virus is by detecting chest X-rays in the patient. Over the past decade's mechine learning technology has developed rapidly and is integrated into CAD systems to provide accurate accuracy. This research was conducted by detecting thoracic radiographs using feature extraction Hu-Moments, Harralic and Histogram and detecting the best accuracy with a classification algorithm to detect the results of COVID-19. The study was conducted by testing the dataset obtained from the Kaggle repository which has images, namely 1281 X-rays of COVID-19, 3270 X-rays Normal, 1656 X-rays of pneumonia, and X-rays of bacteria-pneumonia 3001. In general, this research is included in the Good category because it produces the highest accuracy by the Random forest classification algorithm where the accuracy result is 84% and the standard deviation is 0.015847. In addition, the research also produced Kappa of 0.713. The results of this accuracy are carried out in several stages, namely by feature extraction in the form of hu-moments, Harralic and histogram. In this study, the best results were given by the Random forest algorithm with feature extraction Histogram and Hu-Moment.