Constructing Intelligent Credit Rating Prediction Models by the Application of Financial Variables

In reviewing the history of the Asian financial markets, the financial systems of the nations in Asia suffered great losses and even became credit crisis when the jeopardizations of the heavy wave of the Asian Financial Crisis in 1997, the financial turmoil originated from a US subprime mortgage crisis in the middle of 2007-2009, and the crisis of Dubai debt default in recent days, particularly in the case of banks. It is shown that the negative economic effects of banks are far over those of the general firms, and thus to urgently look reliably forecasting tools providing objective information in assessing the operating performance, financial status, and solvency of banks is needed by the investors. Hence, it is desired to have an early warning system that identifies potential bank failure or high-risk banks through the traits of financial information from an intelligent perspective for surviving in such a continuously changing environment of financial markets. Under this context, the proposal aims to forecast the credit rating by the applications of financial ratios that will be extracted from the publicly financial reports in the Asian banking industry from a well-known BANKSCOPE database. Previously, numerous models have been proposed to deal with credit rating problems, but these models suffer three key drawbacks: (1) Statistical techniques rely on the restrictive assumptions on linear separability, multivariate normality, and independence of the predictive variables; (2) some research results lack explanatory and interpretative power; and (3) too numerous variables are used in establishing the classification model, resulting in high dimensions and complex data, as well as a failure to fully differentiate the significance of attributes and understand the cause-effect relationships among attributes. Therefore, to reconcile these aforementioned shortcomings, the proposal offers two intelligent hybrid models based on artificial intelligence techniques to combine experiential knowledge, feature selection, discretization methods, rule filter, artificial neural networks, and rough set theory. The hybrid models provide comprehensible decision rules as knowledge-based systems for interested parties and offer a suitable method for a classification knowledge discovery of credit rating in the Asian banking industry.[[abstract]]回顧亞洲金融市場歷史裡，歷經1997 年亞洲金融風暴、2007-2009 年美國次貸危機的金 融大海嘯和最近杜拜倒債危機等，這些驚濤巨浪席捲全亞洲，造成很多國家金融體系蒙 受巨大損失，其中以銀行業所受影響最大。以上訊息得知，銀行業受負面影響遠較一般 公司深，突顯投資人對銀行業經營績效、財務狀況和償債能力等資訊掌控之強烈需求， 渴望擁有一套有效早期預警制度，免除遭受巨額投資損失之風險。這樣的概念下，本研 究藉由世界著名BANKSCOPE 金融資料庫蒐集亞洲銀行業財務報表和財務比率資料建 立債信評比預測模型。過去債信評比文獻存在三種主要缺點: (1) 統計技術於資料分配受 到假設條件限制，如線性可分割、多變量常態性和變數獨立性等；不幸地，複雜金融環 境而言，常違反上述假設限制；(2) 研究結果缺乏解釋能力；和(3) 建構解決相關金融 財務問題的模型之影響變數太多，導致高資料量和高維度屬性。為改善以上缺點，本研 究擬提出二個智慧混合式預測模型，結合專業知識、屬性選取方法、資料離散方法、規 則過濾器、類神經網路和粗糙集理論等技術優點，提供有價值的資訊和可理解的決策規 則予利害關係人，做為正確投資或決策規劃之參考