基於模型分群之信用風險評估

Credit Risk Assessment Using Model-Based Clustering Analysis

黃瑞卿(Ruey-Ching Hwang) ； 蕭兆祥(Jhao-Siang Siao) ； 許家豪(Jia-Hao Syu) ； 朱至剛(Chih-Kang Chu)

《中國統計學報》 48卷2期 (2010/06) Pp. 146-160

https://doi.org/10.29973/JCSA.201006.0004

摘要

本文使用基於模型分群(model-based clustering; Fraley and Raftery, 2002)方法，將樣本共變異數矩陣區分爲8個不同的類型，據此建立8個高斯混合模型(Gaussian mixture model)，做爲信用風險評估的分類模型基凖。使用這8個高斯混合模型與Hwang et al.(2010)所提議的5個解釋變數組合，應用至台灣上市櫃公司資料，實證結果顯示這8個高斯混合模型中的VEI模型有最好的分類結果，且該結果比台灣經濟新報的台灣企業信用風險指標(Taiwan Corporate Credit Risk Index; TCRI)有較好的分類表現。

關鍵字

特徵值分解 ； 脆弱模型 ； 高斯混合模型 ； 型Ⅰ誤差 ； 型Ⅱ誤差

並列摘要

Using the result of model-based clustering in Fraley and Raftery (2002), this paper builds 8 different Gaussian mixture models to assess credit risk. In applications, these models using the five explanatory variables considered in Hwang et al. (2010) were used to assess credit risk for companies listed in Taiwan stock market. The empirical result indicates that among those 8 models the VET model had the best classification performance. Comparing the result with that of Taiwan corporate credit risk index provided by Taiwan Economic Journal, our suggested classification method has better performance in the sense of yielding smaller Type Ⅰ error rate.

並列關鍵字

Eigenvalue decomposition ； frailty model ； Gaussian mixture model ； type Ⅰ error ； type Ⅱ error

參考文獻

黃瑞卿、蕭兆祥、鐘惠民(2009)。自我相關結構信用評等之預測。證券市場發展。21，43-70。

Altman, E. I.(1968).Financial ratios, discriminant analysis, and the prediction of corporate bankruptcy.Journal of Finance.23,589-609.

Banfield, J. D.,Raftery, A. E.(1993).Model-based Gaussian and non-Gaussian clustering.Biometrics.49,803-821.

Bensmail, H.,Celeux, G.(1996).Regularized Gaussian discriminant analysis through eigenvalue decomposition.Journal of the American Statistical Association.91,1743-1748.

Campbell, J. Y.,Hilscher, J.,Szilagyi, J.(2008).In search of distress risk.Journal of Finance.63,2899-2939.