Hybrid black-box classification for customer churn prediction with segmented interpretability analysis

Arno De Caigny, Koen W. De Bock, Sam Verboven

Abstract: Highlights•Hybrid Segmented models with blackbox classifiers are proposed.•A large benchmark experiment shows merits of these models for customer churn prediction.•A new visualization for hybrid models is proposed based on SHAP.•A case study showcase the new visualization and compare it with alternatives.