Extreme Risk Modeling: A Regression Tree Approach

Quantifying and managing risks characterized by extreme losses is a key challenge in insurance and risk management. Traditional methods often struggle to handle the heterogeneity and heavy-tailed nature of extreme claims, making accurate pricing and reserving difficult. This talk presents a practical modeling framework that combines regression trees with Extreme Value Theory to improve extreme risk assessment. This approach helps identify key risk factors through data-driven segmentation, allowing for better claim classification and evaluation, particularly for severe and extreme losses.Illustrated through applications in cyber risk and natural disasters, this method provides actionable insights for premium calculations, reserve estimations, and stress testing. The focus will be on practical implementation and how these techniques can support decision-making in risk management.