Transparent Decision Support System for Credit Risk Evaluation: An automated credit approval system

Abstract

The assessment of financial credit risk is an important and challenging research topic in the financial domain. The accurate assessment and prediction of financial credit risk play an essential role both on economics and society. Many banking sectors are looking into an automated decision support system for credit risk evaluation. Consequently, many automated credit risk evaluation systems have been proposed, however most of them lack in explainability needed to justify their decisions. To remedy this, this paper proposes a decision support system named Transparent Decision Support System for Credit Risk Evaluation (TDSSCRE), which extracts concise and justifiable rules from neural networks for credit risk decision. The proposed TDSSCRE tunes and prunes the rules to assist in making a decision support system transparent. Finally this transparent decision support system with concise rules justifies the decision for why applications are granted/rejected with a significant predictive accuracy. The decision system is validated with 3 credit risk datasets using 10 fold cross validation. From the experimental results it is observed that the proposed decision system can produce accurate and justifiable rules for classification. The decision system is also compared against two existing rule extraction models.