Comparative Analysis of Neural Network and Fuzzy Logic Techniques in Credit Risk Evaluation

Comparative Analysis of Neural Network and Fuzzy Logic Techniques in Credit Risk Evaluation

Asogbon Mojisola Grace (Department of Computer Science, Federal University of Technology Akure, Akure, Nigeria) and Samuel Oluwarotimi Williams (Institute of Biomedical and Health Engineering, Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences, China)
Copyright: © 2016 |Pages: 16
DOI: 10.4018/IJIIT.2016010103
OnDemand PDF Download:
$30.00
List Price: $37.50

Abstract

Credit risk evaluation techniques that aid effective decisions in credit lending are of great importance to the financial and banking industries. Such techniques assist credit managers to minimize the risks often associated with wrong decision making. Several techniques have been developed in the time past for credit risk evaluation and these techniques suffer from one form of limitation or the other. Recently, powerful soft computing tools have been proposed for problem solving among which are the neural networks and fuzzy logic. In this study, a neural network based on backpropagation learning algorithm and a fuzzy inference system based on Mamdani model were developed to evaluate the risk level of credit applicants. A comparative analysis of the performances of both systems was carried out and experimental results show that neural network with an overall prediction accuracy of 96.89% performed better than the fuzzy logic method with 94.44%. Finding from this study could provide useful information on how to improve the performance of existing credit risk evaluation systems.
Article Preview

Introduction

The banking Industry play an appreciated role in promoting the economic development of any nation in the world. Primarily, banks focus on credit lending to borrowers in other to generate income, which are later invested into local, national, or international community. For some years now, banks have been experiencing financial crisis in credit lending due to the high level risk associated with improper loan decisions often made by credit officers. This risk includes, loss of principal and interest, disruption to cash flow in the banking system, and increased collection cost, which arises when borrowers fail to pay back acquired credit facility in accordance with the agreed terms of the bank. Several methods have been used in the time past for credit risk evaluation. For instance, the traditional method of granting credit to borrowers is based on judgmental concept using the experience of credit officers and the problems associated with this approach include: high cost of training loan officers; inappropriate decisions; longer period of time required to evaluate a risk; and the possibility of making different decision by different loan officers for the same case (Handzic and Aurum, 2001). To address these problems, methods such as credit scoring, discriminant analysis, logistic regression, and multiple regression were proposed to manage credit risk. However, common limitations of these methods are: the credit scoring methods attempted to correct the biasness of the traditional method but sometimes it misclassify applicants, has the possibility of indirect discrimination, it is not standardized and it varies from one market to another (Crook, 1996) and it does not easily accommodate new changes, discriminant analysis and logistic regression assume multivariate normality and homoscedastic that are often violated in the real world banking data (Giang, 2005; Huang H. et al., 2004), multiple regression require model selection which is based on trial and error process (Leondes, 2005).

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 14: 4 Issues (2018): 1 Released, 3 Forthcoming
Volume 13: 4 Issues (2017)
Volume 12: 4 Issues (2016)
Volume 11: 4 Issues (2015)
Volume 10: 4 Issues (2014)
Volume 9: 4 Issues (2013)
Volume 8: 4 Issues (2012)
Volume 7: 4 Issues (2011)
Volume 6: 4 Issues (2010)
Volume 5: 4 Issues (2009)
Volume 4: 4 Issues (2008)
Volume 3: 4 Issues (2007)
Volume 2: 4 Issues (2006)
Volume 1: 4 Issues (2005)
View Complete Journal Contents Listing