Evaluation of Consumer Credit in Jordanian Banks: A Credit Scoring Approach

Abstract

Loan granting is considered one of the main income sources for banks and financial institutions. Therefore, careful assessment should be taken when deciding to grant credit to potential customers. With the rapid growth of consumer credit and the huge amount of financial data developing effective credit scoring models is very crucial. However, this is not the case in all countries. Emerging financial markets still adopt subjective and intuitive approaches in evaluating credit. This paper intends to introduce and describe a credit scoring model that act as a support decision tool for credit analysts in Jordanian Banks. The models are developed by applying statistical and advanced data mining approaches namely Linear Discriminate Analysis (LDA), Logistic Regression (LR) as statistical models, Neural Networks (NN), Support Vector Machines (SVM) as advanced models. The developed models results are compared in terms of overall accuracy, Type I and II errors. The models are evaluated on real Jordanian credit dataset. All the models performed very well, LR achieved the highest classification accuracy while SVM has the lowest Type I error. In general results are promising and show that credit scoring models can play a very effective role as a support decision tool for Jordanian banks.