Calculating the best cut off point using logistic regression and neural network on credit scoring problem- A case study of a commercial bank

Abstract

Credit scoring is a method used to estimate the probability of default or becoming delinquent of a loan applicant or existing borrower. There are several methods used for scoring, such as traditional statistics models like probit and logistic regression, data mining approaches and also artificial intelligence algorithms. In this paper, two high-usage methods on real data of legal customers of a commercial were used, and also their performance have been compared. It was found that logistic regression as a statistic model can estimate a good econometrics model which is able to calculate the probability of defaulting, and also neural networks is a very high performance black box method which can be used in credit scoring problems. Also the best cut off point in both logistic regression and neural network is calculated by these methods which have minimum errors on the available data. Key words: Credit scoring, logistic regression, goodness of fitness, cut off point, neural network.