Analysis Of Cost Prediction In Medical Insurance Using Modern Regression Models

Abstract

There are different kinds of insurances, but the most saturated is the medical (life) insurance domain. As a vast population invests in health insurance, it is hard to keep track of trends. The ineffective analysis of data results in cost overruns and insurance inequity, making its access difficult. There is a need to critically analyze the insurance data and make an insurance policy that is well adapted to the geographical and financial statuses of the insurers. This study aims to predict suitable medical insurance costs based on the patient's biological and demographic factors by using Machine Learning Regression techniques. Four models are applied on a US-based dataset. Gradient Boosting Regressor, AdaBoost Regressor, Lasso and Elastic Net Regression. Various loss functions were used to extract the best model on different parameters. Overall, the best performance in terms of maximum R2 and minimized loss were given by boosting techniques as compared to the regularization techniques. Proposed system will help organizations to design more public-oriented medical insurance policies which benefit the users and also improve the revenue of the organization.