Abstract

The insurance industry is affected by the COVID-19 pandemic because it contractually covers mortality and health risks. There are many different impacts, some of which balance out the bad ones. Over the past few decades, the health insurance market has been crucial to the overall growth of the Indian insurance sector. This study suggests using COVID-19 patient data to estimate risk using a machine learning-based method. The dataset undergoes a thorough preparation process that involves normalizing it with the Min-Max scaler, handling missing values, and identifying outliers. The RF and LR classifiers are evaluated with respect to measures such as ROC-AUC, F1score, re-call, accuracy, and precision. With an AUC of 0.921 and an accuracy of 89.58%, LR beats RF in the experiments, whereas RF only manages an AUC of 0.504 and an accuracy of 88.19%. The suggested LR model outperforms methods like XGBoost, K-Nearest Neighbors, and Linear SVM in terms of accuracy, according to a comparative study with other models. These results demonstrate how well machine learning approaches may improve risk assessment for life insurance underwriting.