Document Type



This item is available under a Creative Commons License for non-commercial use only


Computer Sciences

Publication Details

Technical Report: SOC-AIG-001-09 (2009)


Consider a bank which wishes to decide whether a credit applicant will obtain credit or not. The bank has to assess if the applicant will be able to redeem the credit. This is done by estimating the probability that the applicant will default prior to the maturity of the credit. To estimate this probability of default it is first necessary to identify criteria which separate the "good" from the "bad" creditors, such as loan amount and age or factors concerning the income of the applicant. The question then arises of how a bank identifies a sufficient number of selective criteria that possess the necessary discriminatory power. As a solution, many traditional binary classification methods have been proposed with varying degrees of success. However, a particular problem with credit scoring is that defaults are only observed for a small subsample of applicants. An imbalance exists between the ratio of non-defaulters to defaulters. This has an adverse effect on the aforementioned binary classification method. Recently one-class classification approaches have been proposed to address the imbalance problem. The purpose of this literature review is three fold: (I) present the reader with an overview of credit scoring; (ii) review existing binary classification approaches; and (iii) introduce and examine one-class classification approaches.