MACHINE LEARNING IN HEALTHCARE MANAGEMENT FOR MEDICAL INSURANCE COST PREDICTION

Machine learning projects have been providing a better patient experience in care services. Healthcare has many issues and the cost of it is an essential indicator for insurance providers. In this context, the purpose of the present paper is to offer a comparison of machine learning approaches for the prediction of American medical insurance cost provided by Kaggle community with 1.338 instances. The focus did not consist of winning any competition, but developing a preliminary investigation of algorithms’ performance assessment. Linear Regression regularizations were compared with more sophisticated algorithms, KNR, SVR, Simple Tree, Random Forest and XGBoost in terms of accuracy with R² and RMSE along with computational time. Linear Regression and its regularizations presented a good accuracy with a five-fold cross validation. However, GridSearchCV selection for best parameters achieved superior performance for more advancedalgorithms, except Support Vector Machine that did not exhibit competitive accuracy. Computational time revealed to be an interesting assessment and depending on the organizational context, simple tree, R² 0.88, would occasionally overcome the others, since it had a competitive computational time comparing to XGBoost and Random Forest, the oneswith the highest accuracy. The present study has contributed on proving machine learning value for health insurance price prediction and the importance of applying comparative performance metrics for the algorithms not only in accuracy, but also in computational time.