Two-step Feature Selection for Predicting Mortality Risk in COVID-19 Patients

Abstract

COVID-19 pandemic is causing serious impact on our society. The whole world is suffering from financial, social, psychological, and other health crisis. One of the various challenges faced is the lack of health and medical facilities around the globe. It is very crucial to properly manage the available resources to save the lives of COVID-19 affected patients. This study proposes an intelligent model to facilitate the hospitals and medical facilities to diagnose which patients are in serious conditions and needs priority health services. The proposed model is based on feature selection-based mechanism, where most dominating features are identified to best discriminate among the serious patients and the less affected patients. We adopted two-step strategy, where filter measure is applied to rank the features according to their relevance in the first step, and Genetic Algorithm is applied with Decision Tree classifier to find the best feature subset in the second step. The results are reported in terms of classification accuracy and the most dominating features are also identified to help the medical practitioners.