FORECASTING HEALTH INSURANCE PAYER INCOME: A COMPARATIVE ANALYSIS OF DECISION TREE AND SVR ALGORITHMS

An insurance company is a type of non-bank financial institution that protects clients from risks and collects premiums over a certain period, these facts provide an overview of the insurance business and highlight its role in the economy, this study evaluated the performance difference between the Decision Tree Regressor and Support Vector Regression (SVR) in predicting insurance payer income. The Decision Tree model demonstrated strong predictive accuracy, achieving a Mean Absolute Error (MAE) of approximately 57 million and an R-squared (R²) value of 0.896, meaning it could explain around 89.6% of the variance in the data. Additionally, the model maintained high consistency, as evidenced by 5-fold cross-validation scores ranging from 0.908 to 0.967, indicating strong generalization and low risk of overfitting. In contrast, the SVR model significantly underperformed. It recorded a much higher MAE of over 237 million and a large Mean Squared Error (MSE), reflecting substantial deviations from the actual values. Its R² score of -0.299 suggests that SVR performed worse than a naive mean predictor, failing to identify meaningful patterns. This poor performance was consistent across all cross-validation folds, which also produced negative R² scores. The SVR model’s inadequacy is likely due to the large scale of the income data and the lack of proper preprocessing, such as normalization, or parameter tuning. Overall, these findings clearly demonstrate that the Decision Tree Regressor is a more suitable, accurate, and stable model for predicting insurance payer income.