Enhancing early detection of COVID-19 with machine learning and blood test results

The medical community has a significant interest in an accurate, reliable, and effective Medical Diagnosis Support System for COVID-19. Compared to the RT-PCR test, clinical database-based COVID-19 diagnosis has been proven to have superior sensitivity and specificity. Recent research has highlighted the benefits of Machine Learning and routine blood testing in the initial screening of COVID-19 patients. The present manuscript proposes a supervised MDSS for COVID-19 prediction that automatically learns from each patient’s clinical data. The objective is to develop and validate a diagnosis support system architecture for COVID-19 prediction based on both qualitative and quantitative blood features. The suggested approach employs statistical techniques to establish a correlation vector of important blood features, and five separate classification techniques (AdaBoost, XGBoost, KNN, Random Forest, and SVM) are then used to evaluate the system. The GridSearch function is employed to identify the best configuration for each generated model, which enhances the accuracy of the final prediction. Additionally, an ensemble ML algorithm combines the predictions from the five best-performing models. Many performance indicators are used to assess the robustness of the proposed method, and the findings demonstrate that the MDSS achieved a high accuracy of 97.52% with a 95% confidence interval of 93.9–100%. These findings indicate better performance than previous work.