Machine Learning Approaches for Medical Insurance Cost Prediction

Abstract

. The question of medical insurance pricing is a significant concern at the contemporary healthcare level, because effective anticipation of medical costs allows healthcare workers to effectively cope with the risk and to develop a fair insurance policy. This paper uses and compares two regression techniques, Linear Regression and Random Forest, with the use of the Kaggle Medical Cost Personal Dataset. The obtained results indicate that the random forest model is infinitely better than the Linear Regression, making errors in predictions lower and attaining a high value of R2 (0.865). The analysis of the feature importance also indicates that the most significant determinants of medical insurance bills are smoking status, BMI, and then age. This indicates that the benefit of machine learning techniques in improving predictive accuracy and determining the most crucial cost drivers, offering quasi-realistic hints in pricing of insurance and in risk control.