|
Abstract:
|
In telecommunication companies, credit risk models are built to help determine (any and how much) mitigation a customer needs to pay for services requested. However, customers who walk away can’t be included in modeling. It might be highly misleading to solely base the credit decisions on behaviors and characteristics of accepted applicants. Reject inference can be used to solve this problem and improve the quality of credit scoring models. In this paper, we will brief introduce the three methods (parceling, fuzzy and hard cutoff) and then discuss how we conducted reject inferencing for a machine learning credit model project.
|
