PREDICTIVE MODEL FOR HEALTH INSURANCE COST USING SELF-ORGANIZING MAPS AND XGBOOST

Abstract

Machine Learning (ML) techniques are gaining more adoption in every sector in order to improve their services. The healthcare industry is not left behind in this development of adopting ML predictive model to increase their efficiency and productivity.The paper developed a predictive healthcare insurance cost Model  using  Self-Organizing Maps (SOM) and XGBoost models.  In this study, two models, SOM and XGBoost were deployed for medical insurance cost prediction using the dataset from KAGGLE’s repository which consists of 1338 instances and 7 predicting parameters. The dataset were preprocessed and thereafter divided into 80% for training and 20% for testing. The comparative result from the prediction showed that the two models achieved impressive outcomes, and whereas the XGBoost model achieved the results of MAE score of 2432.04, MSE of 18030522.49 and RMSE of 4246.24. The SOM model achieved result of MAE score of 3978.29, MSE of 32775593.13 and RMSE of 5724.997216804203. The quantization error of 0.5135462765843376 and topographic error of 0.9730941704035875 generated for SOM model developed. The study concluded that XGBoost outperformed SOM for the insurance predictive model developed and the model is recommended for healthcare sectors to assist in decision making as regard to health insurance cost. More future works can be done using more predicting factors in the dataset and other machine learning algorithms can be employed.