應用自組性演算法建構企業信用評等模型

Abstract

信用評等是金融機構用以評量借款企業償債能力的重要依據。由於國內近年來倒帳事件頻傳，導致銀行對中小企業逾放比居高不下，且「新巴塞爾資本協定」(New Basel Capital Accord)首度明文規定允許銀行使用內部信用評等方法來衡量借款企業之信用風險，因此，金融機構如何建構一套有效的信用評等模型已迫在眉睫。目前中外相關文獻已發展出一些信用評等模型，其中又以無母數方法整合倒傳遞類神經網路之複合式模型(hybrid model)之預測效果較佳，但由於現有的這些方法均無法求得一個數學方程式且網路建構過程複雜，因此金融機構在實務應用上有其困難；此外，還有文獻利用存活分析方法來建構信用評等模型，但此作法常會有資料不符合存活模型前提假設的問題。由於自組性演算法(group method of data handling, GMDH)可發展出一數學方程式，能克服實務應用上的困難且不需任何模型假設，因此本研究應用GMDH及企業內部財務以及非財務變數資料，先建構一個存活期預測模型，預測借款企業有能力還款的時間長度，然後利用此模型之預測結果發展出一個參考指標來判定企業之信用等級。本研究最後利用國內某金融機構近五年向其借款企業之歷史資料驗證了本研究所發展之信用評等模型確實可有效地判別借款企業之信用等級。

Due to the increasing global competition, getting sufficient capital has been a very important issue to companies. For this reason, credit rating is a crucial task for banks or loan companies in the recent decades. Consequently, developing a reliable credit rating model has become an urgent issue for loan companies. Among many studies on credit rating, back-propagation neural networks (BPN) or hybrid model based on non-parameter method and BPN has been proven to have more accurate prediction results as compared to other conventional credit rating techniques. However, the poor explanation ability makes ANN difficult to produce interpretable results. Moreover, constructing hybrid model is very difficult. These drawbacks make BPN and hybrid model have little practical use. Some studies employ survival analysis to construct a credit rating model, but these methods require that data must follow certain distributions. Group method of data handling (GMDH) is a technique which requires no assumptions and can be utilized to develop a mathematical model. Therefore, this study proposes a credit rating model using survival analysis and GMDH. Finally, a real case from a Taiwan’s loan company is utilized to demonstrate the effectiveness of the proposed method.