Revealing Geography-Driven Signals in Zone-Level Claim Frequency Models: An Empirical Study using Environmental and Visual Predictors

2026-04-23

AI Summary

This study addresses the challenge of modeling claim frequency in motor third-party liability insurance, which is often constrained by the lack of granular geographic information in publicly available actuarial data. Leveraging a regional-level framework, the authors enrich traditional actuarial variables with geospatial proxies derived from OpenStreetMap, CORINE land cover data, and Belgian orthophotos, incorporating coordinates, environmental features, and image embeddings extracted via a Vision Transformer. Empirical results demonstrate that the representation of geographic information has a greater impact on predictive performance than model complexity. Even in the absence of individual-level spatial data, multi-source geographic proxies effectively capture risk heterogeneity. The best-performing model combines 5-kilometer-scale environmental features with coordinates, while visual embeddings significantly enhance the accuracy and stability of regularized generalized linear models when environmental features are unavailable.

Technology Category

• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Computer Vision: Remote Sensing / Geospatial AI
• Machine Learning: Calibration & Uncertainty Quantification

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Security and Privacy: Data transparency and provenance

Abstract

Geographic context is often consider relevant to motor insurance risk, yet public actuarial datasets provide limited location identifiers, constraining how this information can be incorporated and evaluated in claim-frequency models. This study examines how geographic information from alternative data sources can be incorporated into actuarial models for Motor Third Party Liability (MTPL) claim prediction under such constraints. Using the BeMTPL97 dataset, we adopt a zone-level modeling framework and evaluate predictive performance on unseen postcodes. Geographic information is introduced through two channels: environmental indicators from OpenStreetMap and CORINE Land Cover, and orthoimagery released by the Belgian National Geographic Institute for academic use. We evaluate the predictive contribution of coordinates, environmental features, and image embeddings across three baseline models: generalized linear models (GLMs), regularized GLMs, and gradient-boosted trees, while raw imagery is modeled using convolutional neural networks. Our results show that augmenting actuarial variables with constructed geographic information improves accuracy. Across experiments, both linear and tree-based models benefit most from combining coordinates with environmental features extracted at 5 km scale, while smaller neighborhoods also improve baseline specifications. Generally, image embeddings do not improve performance when environmental features are available; however, when such features are absent, pretrained vision-transformer embeddings enhance accuracy and stability for regularized GLMs. Our results show that the predictive value of geographic information in zone-level MTPL frequency models depends less on model complexity than on how geography is represented, and illustrate that geographic context can be incorporated despite limited individual-level spatial information.

Problem

• geographic context
• claim frequency
• actuarial modeling
• spatial information
• MTPL

Innovation

• geographic representation
• environmental predictors
• image embeddings
• zone-level modeling
• actuarial prediction