ARIA Prize-Winning Paper Offers Powerful New Tools for Detecting Insurance Fraud

The CAS and the American Risk and Insurance Association (ARIA) have awarded Denisa Banulescu‐Radu and Meryem Yankol‐Schalck the ARIA Prize for their paper, “Practical guideline to efficiently detect insurance fraud in the era of machine learning: A household insurance case,” published in the Journal of Risk and Insurance, Volume 91, Issue 4. The CAS-sponsored prize, established in 1997 and discontinued in 2025, is awarded to the author of a paper published by ARIA that provided the most valuable contribution to casualty actuarial science.

The paper delivers groundbreaking advancements for actuaries working in fraud detection. By applying interpretable machine learning techniques, especially XGBoost enhanced with SHAP values, the study provides a scalable, real-time framework for identifying suspicious claims at first notice of loss (FNOL). Drawing on real French household insurance data, the paper addresses key challenges such as data imbalance, model interpretability, and operational efficiency. Importantly, it bridges cutting-edge data science with actuarial practice, enabling more accurate risk scoring and reducing false positives. Its findings empower actuaries to better prioritize investigations, cut claims-handling costs, and protect honest policyholders—making it a deserving recipient of the ARIA Prize.

Papers eligible for the prize included articles, workshop articles, and invited papers published in the Journal of Risk and Insurance during the preceding year. Papers published in new ARIA journals were also eligible for this award. If no paper was considered eligible in a given year, the award was not made. Papers were judged by a specially appointed committee of the Society, whose decision was final.