Peeking into the Black Box: An Actuarial Case Study for Interpretable Machine Learning

Abstract

This tutorial gives an overview of tools for explaining and interpreting black box machine learning models like boosted trees or deep neural networks. All our methods are illustrated on a publicly available real car insurance data set on claims frequencies.

Keywords: XAI, machine learning, explainability, interpretability, black box models, model-agnostic technique, flashlight, motor insurance, claims frequency

JEL Classification: G22, C52