A Credit Scoring Model of Commercial Banks: the Case of Personal Loans

• Main Authors: Yung-cheng Chen, 陳勇誠
• Other Authors: Chin-Sheng Huang
• Format: Others
• Language: zh-TW
• Published: 2003
• Online Access: http://ndltd.ncl.edu.tw/handle/47599959942212627129

碩士 === 國立雲林科技大學 === 財務金融系碩士班 === 91 === The open up of new banks has strengthened the market competition in Taiwan financial industry. In particular, the higher spread commercial loan sector is the focus of each financial institution. A credit scoring model and a risk management mechanism are essential for a bank to succeed in such a turbulent environment. This study employs Steenackers and Goovaerts’ (1989) as credit scoring model for personal loans. We apply backpropagation network model and discriminant analysis to pattern recognition in the scoring model. Moreover, this study empirically test and compare the performance of rating models. The empirical data, from the personal loan department of a commercial bank in Taiwan (1990-1991), include 1067 loan samples that can be further divided by 794 normal loans samples and 273 default loans samples. This study constructs a Steenackers and Goovaerts’ twelve variables credit model for the personal loan data. The empirical results show: (1) backpropagation network is able to correctly identify in-sample pattern 95.27%, with normal loans and default loans 95.05% and 95.49%, respectively; (2) discriminant analysis is capable to correctly identify in-sample pattern 69.65%, with normal loans and default loans 66.08% and 69.65%, respectively; (3) backpropagation network has out-sample forecasting accuracy 78.19%, with normal loans and default loans 60.77 and 95.60%, respectively.