Credit risk evaluation using neural networks: Emotional versus conventional models

Credit scoring and evaluation is one of the key analytical techniques in credit risk evaluation which has been an active research area in financial risk management. Artificial neural networks (NNs) have been considered to be accurate tools for credit analysis among others in the credit industry. Lately, emotional neural networks (EmNNs) have been suggested and applied successfully for pattern recognition. In this paper we investigate the efficiency of EmNNs and compare their performance to conventional NNs when applied to credit risk evaluation. In total 12 neural networks; based equally on emotional and conventional neural models; are arbitrated under three learning schemes to classify whether a credit application is approved or declined. The learning schemes differ in the ratio of training-to-validation data used during training and testing the neural networks. The emotional and conventional neural models are trained using real world credit application cases from the Australian credit approval datasets which has 690 cases; each case with 14 numerical attributes; based on which an application is accepted or rejected. The performance of the 12 neural networks will be evaluated using certain criteria. Experimental results suggest that both emotional and conventional neural models can be used effectively for credit risk evaluations, however the emotional models outperform their conventional counterparts in decision making speed and accuracy, thus, making them ideal for implementation in fast automatic processing of credit applications. (C) 2011 Elsevier B.V. All rights reserved.