In auto insurance rate-making, traditional approaches typically rely on Generalized Linear Models (GLMs) to estimate risk factor relativities for calculating premium rates across different risk groups. This study proposes an integrated and interpretable analytical framework for evaluating risk factors and territorial classifications using synthetic Usage-Based Insurance (UBI) data. Our approach enhances conventional techniques by integrating Generalized Additive Models (GAMs) for numerical variables, which increases both flexibility and interpretability in individual risk analysis. GAMs generally provide closer agreement with empirical risk patterns. Within the grouped actuarial validation framework considered in this study, GAMs offer enhanced interpretability through flexible nonlinear functional estimation. We also introduce a territorial risk classification framework using interpretable clustering to refine auto insurance pricing in a way that can help better understand the contribution of risk factors. Our aim is to identify territorial clustering patterns based on low-dimensional risk factors. Findings suggest that combining key risk factors with interpretable territorial clustering enhances risk stratification and supports transparent UBI rate-making. The proposed framework offers a practical and interpretable approach for regulatory evaluation and the design of UBI pricing structures.
