Efficient COVID-19 Diagnosis Using Ensemble Learning Model

During the COVID-19 pandemic, new variants of the SARS-CoV-2 virus that are more infectious, virulent, or immunologically evasive have emerged, underscoring the importance of research in diagnosing SARS-CoV-2 virus. In this paper, a predictive model for diagnosing COVID-19 patients using routine blood tests and machine learning techniques is presented. This study implements and evaluates a machine learning model that employs the bagging ensemble learning technique, with Random Forest as the primary classifier. The model has attained an average accuracy of 96.41% and an AUC of 84% on the first dataset. For the second dataset, it has achieved an average accuracy of 98.33% and an AUC of 90.00%. Comparative experimental analysis has shown that our model outperforms various other models, including K-Nearest Neighbors, Extra Tree Classifier, Three-Way Random Forest, Naïve Bayes, and Extreme Gradient Boosting, across multiple performance metrics.