Severity Classification for COVID-19 Infections based on Lasso-Logistic Regression Model
Zainab Hussein Arif, Korhan Cengiz
Abstract
The tremendous growth of the Covid19 epidemic in recent months is devastatingly affecting human civilization. Many different biomarkers are being studied to monitor the patient's health. This might mask the symptoms of various diseases, making it more challenging for a doctor to make a correct diagnosis or prognosis. Therefore, this study aimed to create several classes of prediction methods that can handle situations of varying severity (severe, moderate, and mild). Using machine learning, a Lasso-logistic regression model is developed. To create the Covid-19 clinical dataset, researchers enlisted the help of 78 patients from the Azizia main hospital sector, the Wasit Health Directorate, and the Ministry of Health. The results show that the proposed method is generally accurate to 85.9%. Deaths have been reduced thanks to the established prediction method that enables early detection of patients across three severity levels.