Practice of Data Envelopment Analysis (DEA) in Discriminate Analysis (DA) and it`s usage in ranking of credit customer of Bank Mellat

Abstract

Failure to pay the granted loans in a timely manner is seen as one of the most important challenges faced regarding facilities granting in the banking system of the country which results in outstanding, delayed, and doubtful claims accounts.

One of the suitable solutions to face such challenge is to separate clients with good records from those having bad records prior to grant any loan. Measuring the credit risks of the clients before granting the relevant loan may be considered as a certain solution to reduce the credit risk. The purpose of this study is to present a certain model, which chooses effective features of the client in measuring the credit risk by taking benefit from the elites’ judgment and applies data coverage analysis in discriminate analysis (DA) to identify clients with good records from those with bad records. Such presented model was tested by using dataset of bank’s clients and through considering 20 features from 500 borrowers. Through utilization of Delphi method, eight effective features of the clients in predicting the credit risk were identified, and further, based on the number of the outstanding installments, the clients were classified into two groups of having good records (G1) and bad records (G2). Later on, through placing the data in DA model,,  and d weights relevant to the super-planes separating the two foregoing groups were obtained. Based on such weights, it is possible to determine that the new data shall be placed in which group. Considering the high precision of prediction of the model in classification of the clients into two groups with good and bad records prior to grant loan, such model may be used as a decision-assisting system to help those charge of facilities in the banks. Therefore, not only loans will not be paid to those clients with bad records, but also certain measures (inter alia, obtaining sufficient pledge) may be taken to grant loans to clients with good and/or bad records. Hence, in addition to take a practical step towards reducing the bank’s claims, the level of satisfaction of the clients with good records may also be increased.