應用集群分析與決策樹建構中小企業風險評估模型

Abstract

近年來經濟環境變遷快速，逾收款問題嚴重，致使銀行或金融機構之放款風險大為增高。台灣金融機構或銀行放款的對象大多是中小企業，但由於大部分的中小企業財務報表不夠透明，使金融機構或銀行放款給中小企業的風險相當高。巴塞爾銀行監理委員會於2004年6月底公布了新巴塞爾資本協定，在其協定中增列內部評等法，允許金融機構自行做風險評估來衡量借款客戶的風險。在風險評估的相關文獻中，所使用的分析工具從簡單的敘述統計、多變量方法，到類神經網路等來建構企業之風險評估模型，然而這些文獻大多是以上市上櫃公司做為研究對象，針對中小企業風險評估方面之研究則相當罕見，故本研究利用集群分析(Cluster analysis)與決策樹(Decision tree)發展出一套中小企業風險評估模型，不僅有易瞭解的規則，且此風險評估模型也具有相當不錯的預測能力，易於金融界使用。本研究之實例說明部分是利用國內某金融機構所提供之中小企業借款資料來驗證本研究之模型的可行性與有效性。

【關鍵詞】風險評估、集群分析、決策樹

Because the economic environment changes rapidly in recent years, bad accounts of banks or financial institutions have increased. Consequently, the financial risk of banks or financial institutions also increases rapidly. Moreover, about 90% of the businesses in Taiwan are small and medium-sized enterprises, and the financial reports of these enterprises are not available. This also increases the risk of Taiwanese banks or financial institutions which loan the money to the small and medium-sized enterprises. In order to decrease the financial risk, Taiwanese government has required all banks and financial institutions in Taiwan to develop their own internal risk assessment model according to the New Basel Capital Accord (Basel II) in 2006. For this reason, constructing a reliable risk assessment model has become an important issue for banks and financial institutions in Taiwan. In the previous studies of risk assessment, tools often used for constructing the risk assessment models are some statistical methods or neural networks. These models mainly are for publicly traded companies. Studies of developing a risk assessment model for small and medium-sized enterprises are rarely seen. Hence, this study develops a risk assessment model for Taiwanese small and medium-sized enterprises using cluster analysis and decision tree. The proposed method is easy to implement and has good forecasting ability. Finally, a real case from a Taiwanese loan company is utilized to demonstrate the effectiveness of the proposed procedure.

Key Words: Risk Assessment model, Cluster Analysis, Decision Tree