結合案例式推理和基因演算法建構台灣中小企業信用評等決策模型

Abstract

隨著金融環境不斷變化，金融風險亦隨之提高，國際清算銀行巴塞爾銀行監理委員會於2004年6月公布新版巴塞爾資本協定，其中規定應計提之最低資本要求中，鼓勵銀行內部自行建立信用評等模型，來衡量向銀行借款客戶的風險。行政院金融監督管理委員會也在2006年要求本國金融機構開始實施此協定作業。現有中外文獻所提出之信用評等模型在資料取得不易的限制下，大多是以上市和上櫃公司為研究對象，然而台灣的產業結構以中小企業為主體，其特性與上市和上櫃公司有所不同，使得現有中、外文獻所建立之信用風險評等模型可能不適用於評估台灣中小企業之信用等級。因此，本論文主要目的是發展一套簡便且準確之評估台灣中小企業借款者信用等級之決策模型，以利金融機構根據申請貸款之中小企業的各項財務及非財務資料快速決定其信用等級，金融機構再根據借款企業之信用等級制定適當之放款決策，以降低放款風險。本研究應用案例式推理方法，針對新借款中小企業資料，從案例庫中搜尋與新案例相似之舊借款中小企業案例，從舊案例信用等級來判定新借款中小企業之信用等級。案例式推理屬性權重之訂定非常重要，在案例式推理的相似度比對過程中，本研究應用基因演算法搜尋每個屬性的最佳權重，以提升案例式推理決定借款企業信用等級的正確性。本研究最後以台灣中小企業借款客戶的實際歷史資料，證實本研究所建構之中小企業信用評等模型確實有效。

Due to the continuous changes in the global financial environment, financial risk increases rapidly for banks or financial institution. In June, 2004, Bank for International Settlements Basel Committee on Banking Supervision announced the New Basel Capital Accord (Basel II). In order to reduce the financial risk, banks are encouraged to develop their own credit rating model to measure the risks of their borrowers. Financial Supervisory Commission has also required domestic financial institutions to bring Basel II into force since 2006. Under the restrictions of confidential financial data from enterprise borrowers, many studies on credit rating models utilized from publicly traded companies instead of small-and-medium sized enterprises. Small-and-medium sized enterprises play important roles in Taiwan industry. It may be inappropriate to use existing credit rating models to measure the risk of small-and-medium sized enterprises in Taiwan. Thus, the main purpose of this study is to develop an easy and accurate credit rating model for small-and-medium sized enterprises in Taiwan. By utilizing the proposed method, banks or financial institutions can give appropriate credit ratings promptly for borrowers of the small-and-medium sized enterprises according to their financial and non-financial data. Financial institutions can make appropriate loan decision according to the credit ratings of enterprise borrowers to reduce their loan risks. This study applies Case Based Reasoning (CBR) to determine the credit ratings of new borrowers of small-and-medium sized enterprises by referring to the credit ratings of similar cases from database. It is very important to set the weights of the attributes in CBR. In the process of similarity matching in CBR, Genetic Algorithms is utilized in this study to determine optimal weights of the attributes of indexing to increase the predictive accuracy of credit rating. Finally, real data of small-and-medium sized enterprise borrowers provided by a financial institution in Taiwan to verify the effectiveness of the proposed method.