Assessing Default Probability of EN Bank Legal Customers using Support Vector Machine Method

According to current development in credit allocation and recent economic crises, planning for identification of credit risk has found special importance for investors, banks, shareholders and financial analysts, so that they are able to make proper decisions. Although credit loss is a common cost in banking industry, however, increase in this loss might affect the bank performance. Therefore, there is a strong need to reassess current approaches in risk evaluation of each loan and default rate of loan portfolios. Banks usually have their own internal validation models for loan risk measurement but these approaches are inappropriate and utilize simple mathematical approaches based on incomplete premises. In this paper, we have tried to estimate the possibility of default for legal customers using 20 financial ratios for 200 healthy and 200 unhealthy companies receiving civil participation facilities from Eghtesad Novin (EN) Bank in 2009 and 2010 and 4 approaches for choosing financial ratios including remarks from credit experts of Raah Eghtesad Novin Co., Altman, comparison between averages and choosing correlation attribute. Results show that Support Vector Machine approach can differentiate between healthy and unhealthy companies with average accuracy of 84.63% using all chosen ratios.