In the insurance industry, it is a foundational task to forecast the insurance claims with a very high accuracy for the risk assessment, reserve management, and the premium calculation. The linear regression models have historically dominated in insurance because of their simple nature and interpretability; however, they often fall short in apprehending the nonlinear relations that are available in the complete insurance data sets. Polynomial regression is the extension of linear regression that allows for higher-order interactions among features and offers a practical center ground between simple linear models and complex machine learning algorithms. This literature investigates the application of polynomial regression for insurance claims forecasting by using a real-world auto insurance dataset. We inspect the model’s predictive power, interpretability, overfitting challenges, and how it associates with tree-based ensemble models like random forest and gradient boosting. The results disclose that polynomial regression achieves noteworthy improvements over linear models while maintaining the transparency, which makes this a practical model for actuaries and data scientists.
Jain et al. (Fri,) studied this question.