mSHAP: SHAP Values for Two-Part Models

Two-part models are important to and used throughout insurance actuarial science. Since is required for registering a car, obtaining mortgage, participating in certain businesses, it especially that the price policies fair non-discriminatory. Black box can make very difficult know which covariates influencing results, resulting model risk bias. SHAP (SHapley Additive exPlanations) values enable interpretation of various black models, but little progress has been made two-part models. In this paper, we propose mSHAP (or multiplicative SHAP), method computing using individual This will allow predictions be explained at an observation level. After developing mSHAP, perform in-depth simulation study. Although kernelSHAP algorithm also capable approximate model, comparison with our demonstrates exponentially faster. Ultimately, apply ratemaking personal auto property damage coverage. Additionally, R package (mshap) available easily implement wide variety applications.