Performance Comparison of Multiple Supervised Machine Learning Algorithms for COVID-19 Mortality Prediction

ABSTRACT

The coronavirus disease (COVID-19) has wreaked havoc on populations around the world. Every day, thousands of people are dying as a result of this lethal virus. Patients with pre- existing conditions, as well as the elderly, are more susceptible to the disease. Artificial intelligence can play a vital role to track patient health conditions using various parameters. It assists in determining how to best handle certain patients in order to save their lives. The various parameters of a patient's health condition may have a significant impact on the outcome. Various artificial intelligence strategies are a blessing in minimizing the loss from COVID-19. This paper focuses on predicting the potential outcome of a patient using the COVID-19 dataset obtained from John Hopkins University of infected patients, which will help minimizing the death toll of COVID-19 disease. In this study, the performance of various machine learning models is compared for predicting COVID-19-affected patient's mortality using Logistic Regression, Support Vector Machine, K Nearest Neighbor, Decision Tree and Gaussian Naive Bayes. Finally, the best model for hyper parameter tuning was chosen from the comparative section. After hyper parameter optimization, a maximum accuracy of 95 percent and an F1 score of 89 percent using the K Nearest Neighbor algorithm was achieved. © 2022 IEEE.