Supervised Machine Learning Models for COVID-19 Prediction

Abstract

The COVID-19 pandemic had a profound impact on global public health, resulting in millions of deaths worldwide. Understanding the factors influencing patient survival outcomes is crucial.This study conducts a comparative analysis of various supervised machine learning models to predict COVID-19 survivors. The dataset sourced from Kaggle repository containing 373 records; however, only 74 records were selected for analysis due to missing data in several feature variables. The outliers were addressed using the Z-Score method, while missing values were imputed using Multiple Imputation by Chained Equations (MICE).We partitioned the dataset into two distinct subsets: 80% (59 data points) for training and 20% (15 data points) for testing. Supervised classification models, including Support Vector Machine, Random Forest, Naive Bayes, Logistic Regression, Decision Tree, and Artificial Neural Network, were employed. The results indicated that the Random Forest model outperformed the others in predicting COVID-19 survivors, with an accuracy of 0.97±0.06, followed by Naive Bayes with an accuracy of 0.75±0.12. This findings demonstrate that Oxygen levels and Age emerge as strong predictors of COVID-19 severity; thus guiding patient outcomes and healthcare services.