Customer determinants of used auto loan churn: comparing predictive performance using machine learning techniques

This paper addresses the use of a customer character model as a determinant of used auto loan churn among a unique population of subprime borrowers. The customer character model (i.e restricted model) is compared to a full model consisting of the 4 Cs of capacity, collateral, credit, and character of churn prediction. The results reveal that there is a difference between the full model and the customer character model. Additionally, different supervised classification methods, such as logistic regression (LR), linear discriminant analysis (LDA), decision trees (DTs), and random forests (RFs), are applied and compared in terms of multiple predictive performance measures. The RF classification measures report the strongest performance. Additionally, different classification methods suggest the importance of different customer character variables. Therefore, from a practical perspective, effective borrower character screening is recommended to determine customer profiles more accurately for the purposes of target marketing and customer retention. This study also deepens understanding of subprime credit markets and reveals additional insights to credit screening using machine learning techniques.