题名 |
導入資料採礦技術於中小企業營造業信用風險模型之建置 |
并列篇名 |
Establishment of Credit Risks Model for the Construction Industry of the SMEs with Data Mining Techniques |
DOI |
10.6338/JDA.200910_4(5).0004 |
作者 |
謝欣芸(Shin-Yun Hsieh);胡麟(Lin Hu);蔡明熹(Ming-Hsi Tsai) |
关键词 |
新巴塞爾資本協定 ; 信用風險模型 ; 羅吉斯迴歸 ; 資料採礦 ; The New Basel Capital Accord ; Credit Risk Model ; Logistic Regression ; Data Mining |
期刊名称 |
Journal of Data Analysis |
卷期/出版年月 |
4卷5期(2009 / 10 / 01) |
页次 |
83 - 112 |
内容语文 |
繁體中文 |
中文摘要 |
為了符合國際清算銀行在2006年通過的新巴賽爾資本協定,且有鑑於近年來整體經濟環境欠佳,銀行業者面對外部的規定以及內部的需求,積極地尋求信用風險模型的建置方法,希望將整個融資的評等過程系統化以提高對信用風險的控管。本研究希望利用92至94年未上市上櫃中小企業之營造業的資料,依循新巴賽爾資本協定之規定並配合資料採礦的技術,擬出一套信用風險模型建置與評估的標準流程,其中包含企業違約機率模型以及信用評等系統的建置,前者能預測出授信戶的違約情形以及違約機率;後者則是能利用前者的分析結果將授信戶分成數個不同的等級,藉此區別授信戶是否屬於具有高度風險的違約授信戶,期待能提供銀行業者作為因應新巴賽爾協定中內部評等法的建置,以及中小企業的融資業務上內部風險管理的需求一個參考的依據。研究結果共選出5個變數作為企業違約機率模型建立之依據,訓練資料以及原始資料的AUC分別為0.799以及0.773,表示模型能有效的預測違約機率並判別出違約授信戶以及非違約授信戶。接著,經過回顧測試與係數拔靴測試,證實本研究的模型具有一定的穩定性。另外,透過信用評等系統將所有授信戶分為8個評等等級,並藉由等級同質性檢定以及敏感度分析的測試,可以驗證出本研究之評等系統具有將不同違約程度的授信戶正確歸類之能力。最後,經由轉移矩陣可以發現,整體而言,營造業在2003年到2005年間的表現有逐漸好轉的趨勢,與營造業實際發展情形相互比較之下,也確實得到相互吻合的結論。 |
英文摘要 |
In order to conform to the New Basel Capital Accord passing in 2006 by the Bank for International Settlements and due to the unsatisfactory with the whole economic environments in the recent years, the banking industry have positively seeking for the way to establish the credit risks model with the exterior regulations and the internal demands. The banking industry hope to systematize the process of scoring in the loan business to enhance the control of credit risks.According to the regulations of the New Basel Capital Accord and the standard process of data mining, I hope to take advantage of the non-listed data of the construction industry of the SMEs from 2003 to 2005 to perform the process of the establishment and evaluation of the credit risks model which includes the default risks model of companies and the credit rating system. The default risks model can predict whether a company will default or not and its default rate. Besides, the credit rating system can give different grade to each company based on the result from the default risks model to discriminate those who have high risks of default. Look forward to providing the banking industry as a result of the applications of the Internal-Rating Based Approach of the New Basel Accord as well as becoming a reference basis of the loan business in the SMEs.The result reveals that five variables were selected as the default probability model basis. The AUC for the training data was 0.799 and for the raw data was 0.773, which represents that the model can effectively predict the probability of default and determine which companies would default. After a series of test, the model in this study has certain stability. In addition, the credit rating system will divide all companies into 8 grades. By homogeneity test and sensitive analysis, we can verify the rating system has the ability to correctly classify different levels of credit default companies. Finally, through the transformation matrix we can find the construction industry have been improvement in the performance of the trends in 2003 and 2005. This result is match for the actual development in construction industry as well. |
主题分类 |
基礎與應用科學 >
資訊科學 基礎與應用科學 > 統計 社會科學 > 管理學 |
参考文献 |
|
被引用次数 |