Improving Convolutional Neural Networks’ Accuracy in Covid-19 Detection Using Support Vector Machine

Abstract

Since December 2019, the coronavirus (COVID-19) pandemic spread in all countries and put health systems under tremendous pressure. Massive efforts have been conducted to find ways to determine the infected patients quickly. Therefore, intelligent systems empowered with Machine Learning and Deep Learning have been utilized in detecting several diseases (especially COVID-19). The systems examine chest x-rays of the suspected patient to decide whether it is a COVID-19 case. This paper evaluates three DL models of Convolutional Neural Networks (CCN): GoogleNet, AlexNet, and VGG16 on COVID-19. The evaluation is based on and without using a Support Vector Machine (SVM) (ML algorithm). To study the robustness of the proposal, we evaluate the following metrics: Accuracy, Precision, Specificity, Sensitivity, and F-measure. The findings demonstrate models empowered SVM superiority in classifying COVID-19 patients perfectly.