Geographic data transforms motor insurance risk models.

Geographic data transforms motor insurance risk models. Researchers analyzed how environmental indicators and satellite imagery improve claim-frequency predictions for motor third-party liability insurance. Using the BeMTPL97 dataset, they tested geographic information from OpenStreetMap, land cover data, and aerial photos across multiple modeling approaches including machine learning algorithms. Results demonstrate that combining coordinates with environmental features significantly enhances prediction accuracy. Notably, image embeddings proved valuable only when traditional geographic data was unavailable. The study reveals that geographic representation matters more than model complexity when incorporating spatial context into actuarial frameworks.