Machine learning-based automation of COVID-19 screening using clinical dataset
Fast and effective screening of COVID-19 patients can help decrease its mortality rate. In this study, we employed machine learning methods on clinical datasets for automated COVID-19 screening. The model configuration with a decision tree as the classifier, forward selection as a feature selection method, and six features has 91% precision, 85% accuracy, 92% sensitivity, and 91% F1-score. We also determined the following relevant features identified by the employed feature selection technique that helped screen COVID-19 patients: lymphocytes, chest X-ray label, BMI, PCT, eGFR result, and comorbidity chronic lung disease. This study demonstrates the potential of machine learning models trained on clinical data in classifying COVID-19 patients to help facilitate the screening procedure.