Abstract:

Coronavirus (COVID-19) disease, which first appeared in December 2019 and caused a worldwide outbreak; is described as a viral disease caused by acute respiratory syndrome SARS-CoV-2. The current RRT-PCR test is used to detect COVID-19 disease. Due to long return time of this test, about 15-20% false-negative rates and expensive equipment, the detection method with the values of routine blood analyses can be considered as a faster and cheaper alternative. In this study, COVID-19 was tried to be detected by using Deep Neural Networks (DNN), one of the routine blood tests. Because there is class imbalance in the used data set, class imbalance has been eliminated by resampling methods and the performance of used algorithms has been evaluated. While resampling, SMOTE, ADASYN, Geometric SMOTE, Random UnderSampler, Random OverSampler algorithms were used. As a result of established model, the most successful result was obtained with the Random OverSampler algorithm, with an accuracy of 0.985 and an F1-score of 0.99. In addition, an application has been developed using PyQt to make predictions for new data to be entered and the contributions of used attributes to the model were determined and explained with the SHapley Additive Explanations (SHAP) technique.

Keywords: