Predictive Modeling for Property Insurance Premium Estimation Using Machine Learning Algorithms

Accurate premium estimation is a cornerstone of effective property insurance, ensuring both financial stability for insurers and fair pricing for policyholders. Traditional actuarial models, though foundational, often fall short when dealing with high-dimensional, nonlinear insurance data. This study presents a machine learning-based approach for accurately predicting property insurance premiums using the Extreme Gradient Boosting (XGBoost) model. Leveraging a comprehensive home insurance dataset with 42 attributes, the proposed methodology involves data preprocessing, feature engineering, training the model, and evaluating its performance using conventional regression measures. The XGBoost model demonstrated superior predictive performance, achieving RMSE = 52.99, MSE = 28.07, and MAE = 30.64 is the calculated value for the mean squared error. These results indicate its strong capability in modeling complex relationships and estimating premiums with high accuracy. Comparative analysis against the Linear Regression model further highlights the effectiveness of the proposed XGBoost model, making it a robust and scalable solution for real-world property insurance pricing systems.