The Study by Using SVM and Logit Model to Predict the Credit Risk of Cash Cards

• Main Authors: Tung-fu Hsieh, 謝東福
• Other Authors: Weissor Shiue
• Format: Others
• Language: zh-TW
• Published: 2008
• Online Access: http://ndltd.ncl.edu.tw/handle/v9j3f5

碩士 === 國立高雄第一科技大學 === 財務管理所 === 96 === Abstract “The New Basel Capital Accord” requests that the bank should establish the system of the risk measurement and estimate the risk to increase the security and the completeness of the financial institutes. The crisis of the credit cards and the cash cards which increased the ratio of non-performing loans make the issued bank have a large amount of bad debts that seriously influences the profit of the bank and reveals the problem of the risk management of the bank. Therefore, how to build a model of the credit risk measurement of the cash cards to decrease and prevent the loss of the non-performing loans is an important subject. The SVM is used to predict the credit default risk of the cash cards and compare with the Logit model. In order to use the SVM to raise the predictive effect and find the risk factors influencing the probability of default of the customers. Therefore, it is necessary to establish the predict models of the risk management of the cash cards and reduce the ratio of the bad debts. In empirical results, according to the SVM method, the correct ratio of the training set and the testing set are 73.61% and 70.56% respectively. And the Logit model correctly predicts 69.86% in a training set and 66.11% in a testing set. In comparison, the predictive ability of the SVM method is better than that of the Logit method. Besides, the times of the credit inquiry in JCIC, gender, and whether to use the revolving loans are the most significant factors to affect the credit risk.