Paper: arXiv 2604.21893

Authors: Sherly Alfonso-Sánchez, Cristián Bravo, Kristina G. Stankova

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.

Complexity vs Empirical Score

  • Math Complexity: 3.5/10
  • Empirical Rigor: 7.5/10
  • Quadrant: Street Traders — practical and empirical, lighter on theory

Why this score: This paper presents a strong empirical evaluation of alternative spatial data (OpenStreetMap, land cover, orthoimagery) for actuarial claim-frequency modeling on the BeMTPL97 dataset. While it relies on established ML architectures (GLMs, GBDT, CNNs, Transformers) rather than novel mathematical proofs, its rigorous zone-level validation and practical focus on feature representation offer high value to applied quantitative risk modeling.

Research Flowchart

  flowchart TD
    A[Research Goal: Predict MTPL Claim Frequency via Geographic Data under Spatial Constraints] --> B[Data Inputs: BeMTPL97 Actuarial Data, OpenStreetMap/CORINE Features, Orthoimagery]
    B --> C[Feature Engineering: Extract Spatial Coordinates, Environmental Features @ 5km, & Vision Transformer Image Embeddings]
    C --> D[Model Training & Evaluation: Test GLMs, Regularized GLMs, Boosted Trees, and CNNs on Unseen Postcodes]
    D --> E[Key Finding 1: Combining Coordinates + 5km Environmental Features Yields Best Accuracy Across Models]
    D --> F[Key Finding 2: Image Embeddings Help Regularized GLMs Only When Environmental Features Are Absent]
    E & F --> G[Conclusion: Geographic Representation Matters More Than Model Complexity for Zone-Level Predictions]