Predictive modeling of COVID-19 hospitalization using twenty machine learning classification algorithms on cohort data

Abstract

Introduction: The global COVID-19 pandemic has led to a health crisis, emphasizing the need to identify high-risk patients for effective resource allocation and prioritized hospitalization. Previous studies have been limited in their use of algorithms and variables, while this research expands to include lifestyle factors and optimizes hyperparameters for twenty machine learning algorithms, enhancing prediction accuracy and identifying key predictors. Material and Methods: In this cross-sectional study, we analyzed data from 207 COVID-19 patients. The Boruta algorithm was used to select the best features for twenty classification algorithms, and RandomizedSearchCV was utilized to optimize hyperparameters. The models were evaluated using performance metrics such as accuracy, f-measure, and area under the curve (AUC). Results: The study identified eight key predictors of COVID-19 hospitalization, which include gamma-glutamyl transpeptidase, alkaline phosphatase, diagnosis by CT scan, mean platelet volume, mean corpuscular volume, fasting blood sugar, red blood cell count, and mean corpuscular hemoglobin concentration. By optimizing the hyperparameters of twenty machine learning algorithms, the accuracy and AUC were improved. With an outstanding AUC of 81.25, the XGBClassifier model exhibited superior performance. Conclusion: The findings of this study can assist clinicians in allocating resources effectively and improving patient care. Additionally, this approach can aid healthcare researchers in leveraging artificial intelligence to manage diseases.