在國際財務報導準則IFRSs下以Crisp-DM流程建立財務風險管理平台

Under the International Financial Reporting Standards (IFRSs), Using Crisp-DM Procedure to Build up a Financial Risk Management System.

陳維毅(Wei-Yi Chen) ； 邱俊誠(Chun-Cheng Chiu)

《Journal of Data Analysis》 8卷3期 (2013/06) Pp. 119-137

https://doi.org/10.6338/JDA.201306_8(3).0006

摘要

國際財務報導準則（International Financial Reporting Standards，簡稱IFRSs）已是會計資訊邁入全球化之重要橋樑及共通語言，其中IFRS7「金融工具：揭露」將所有金融工具之揭露彙整成同一號準則，使報表使用者可獲得一致的風險揭露資訊，提高財務報表之透明度及可比較性，對於報表使用者利用財務報表評估企業財務績效及相關風險影響甚鉅；另外，基於Basel II之下的風險模型亦趨成熟，希冀能導入在Basel II下建立風險模型之經驗，經Crisp-DM流程建立模型，建構在IFRSs下企業財務風險管理之標準流程。本研究嘗試以資料採礦方法與雲端技術建構企業財務風險模型，使用R程式將流程建構於雲端平台，提高建立模型的便捷性及效率。本研究以上市櫃公司之財務報表及公司治理資料，模型建構時採用羅吉斯迴歸、類神經網路、決策樹、隨機森林及支援向量機上述統計方法，藉由比較多種統計方法預測出之模型正確率與反查率，最終選定羅吉斯迴歸模型為最終模型，並作評估及驗證，發現預測能力穩定，整體正確率及反查率皆達八成以上，確實能在實際業務中加以應用。

關鍵字

國際財務報導準則 ； 資料採礦 ； 雲端運算 ； 財務風險 ； 羅吉斯迴歸 ； 類神經網路 ； 支援向量機 ； 決策樹 ； 隨機森林

並列摘要

IFRSs is an important bridge and a common language which let accounting information step into globalization, among which ”IFRS 7 Financial Instruments: Disclosures”, which compiled the disclosure of all financial instruments into the same criteria. Let users of financial statements get consistent risk disclosure information to enhance the transparency and availability of financial statements and impact enormously for users of financial statements use financial statements to evaluate corporate financial performance and associated risk. In addition, based on the risk model under the Basel II becomes maturing, hoping to be able to import the experience of setting up a risk model under Basel II. Through Crisp-DM process to build up models, and construct the standard process of corporate financial risk management under IFRSs. This study tries to construct a corporate financial risk models by use data mining methods and cloud technology. Use the R program to construct the process in the cloud platform to improve the convenience and efficiency of build up models. In this study, use the financial statements and corporate governance data of listed companies, using logistic regression modeling, neural networks, decision trees, random forests and support vector machine. By compare those statistical methods accuracy and recall. Finally, select logistic regression model. By models assessment and verification, found that the stability of the predictive ability. Accuracy and recall are more than 80%, shows can indeed be applied in actual business.

並列關鍵字

IFRSs ； Data Mining ； Cloud Computing ； Financial Risk ； Logistic Regression ； Neural Network ； Support Vector Machine ； Decision Tree ； Random Forests

參考文獻

方匡南、吳見彬、朱建平、謝邦昌(2011)。隨機森林方法研究綜述。Statistics & Information Forum。26(3)

方順逸(2011)。IFRS 7號－金融工具：揭露公報影響試析。IFRS專刊。80-89。

林真真(2007)。統計分析與應用手冊－使用R軟體。台北市:文魁圖書。

()。

金融監督管理委員會（2004）新巴塞爾資本協定中文版