Corporate Default Risk Evaluation Using Support Vector Machines
| Autor: | Yung-Hsin Lee, 李永新 |
|---|---|
| Rok vydání: | 2007 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 95 The purpose of this paper is adopting Support Vector Machines (SVM) to evaluate corporate default risk with financial information. In order to test the efficiency of SVM, this research uses Logit and Z-Score (Altman, 1968) at the same time. The data comprises listed companies in Taiwan Security Exchange Corporation (TSEC) and Over the Counter (OTC) from 1995 to 2005 that they have ever been listed in “Full Delivery” and some normal companies to match with. Because financial variables are used to detect the company’s default, we should not overlook the problems of window dressing and industry difference in Accounting Standard. Thus, we add two non-financial variables default distance (DD value, calculating from option pricing model) and industry dummy variable in the model to reduce the above problems. The results were obtained as following: 1. SVM is better than Logit and Z-Score in prediction. 2. The explanatory variable affects the correct rate and stability crucially. 3. The over-fitting existed in both SVM and Logit, but it can be eliminated obviously in SVM as the explanatory variables were selected efficiently. 4. The prediction is trustworthy in two years. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|