整合KMV模型、約略集合及隨機森林應用於企業信用評等之研究

A study on integrated KMV model, rough set and random forest for firm’s credit ratings.

許智宇(Chih-Yu Hsu)

指導教授 ： 林鳳儀

國立臺北科技大學/管理學院/商業自動化與管理研究所/碩士(2010年)

https://doi.org/10.6841/NTUT.2010.00222

摘要

近年來許多企業雖然擁有較高的信用等級，卻遭受整併或倒閉，因此提供一個有效的信用評等模型是個重要的議題。為解決此問題，本研究利用整合隨機森林與約略集合進行兩階段分類模式的企業信用評等評估模型。此外，在探討信用評等的衡量指標上，本研究參考過去文獻與研究所使用財務性變數之外，亦加入KMV模型的違約距離變數與公司治理等變數建構模型，希望能藉由更完整多元的資訊，來幫助企業本身評估其信用評等，並做出正確的決策。本研究經由理論與文獻的探討，建立了新的信用評等模式，在經過實證的結果發現，經由隨機森林方法針對所考量之衡量信用評等指標進行分析，得知企業的信用評等，除了受到傳統財務變數的影響外亦受到違約距離與公司治理等變數的影響。其中，違約距離在影響信用評等分類結果的重要性高於公司治理變數。再者，有關整合隨機森林與約略集合方法所建構之信用評等模式亦能確實提升信用評等的準確率之外，透過約略集合導出的決策規則亦可提供授信人員作為企業信用評等決策依據。

關鍵字

信用評等 ； KMV模型 ； 隨機森林 ； 約略集合

並列摘要

The development of the firm’s credit rating prediction model has attracted lots of research interests in academic and business community. The objective of this proposed study is to investigate the performance of firm’s credit ratings with random forest and rough set technique. In addition to traditional financial indicators, KMV and corporate governance variables are also included in this model. As the results reveal, we find out traditional financial indicators, default distance and corporate governance variables significantly influence the diagnostic accuracy of firm’s credit ratings by applying our proposed approach. Moreover, our present study indicates that the proposed integrated approach predicts more accurate than solely adopting rough set technique. In the other words, by adopting random forest approach to come out with good initial estimation, rough set approach might takes longer time to achieve accurate results. Finally, we believe the credit rating rules we had summarized in this study could further assist investors’ decision making.

並列關鍵字

Credit ratings ； KMV model ； Random forest ； Rough set

參考文獻

[1] 葉銀華、柯承恩、蘇裕惠，最終控制者的股權與財富剝奪傾向之研究，經濟論文，第31 卷第3 期，2003。

[9] 施人英、陳文華、吳壽山，探討支持向量機器在發行人信用評等分類模式之應用，資訊管理學報 第十四卷 第三期，2007。

[1] Al-Qaheri. H, Zamoon S. and A. Hassanien, “Rough set generating prediction rules for stock price movement,” Computer and simulation, 2008, pp.111-116.

[2] Altman, E.I., “Financial ratios, discriminant analysis and the prediction of corporate bankruptcy,”The Journal of Finance, vol.123, no.4, 1968, pp.589-609.

[3] Angelini.E, Tollo G. and Roli A., “A neural network approach for credit risk evaluation,” The Quarterly Review of Economics and Finance, vol.48, 2008, pp.733–755.