應用資料探勘技術建置預測企業財務危機之模式

Application of data mining techniques build models to predict corporate financial crisis

林博隆(Bo-Long Lin)

指導教授 ： 謝淑玲

嶺東科技大學/資訊學院/資訊科技系碩士班/碩士(2016年)

摘要

企業財務危機問題一直是許多學者探討的問題， 本研究採用台灣經濟新報資料庫蒐集2003年5月至2013年4月符合本研究定義之財務公司，採用以1:2配對抽樣之方式選取樣本比例，52間發生財務危機公司配對104正常公司，財務性變數同時納入學者Altman(1968)及Ohlson(1980)共19個財務變數，進行分析支持向量機、C5.0決策樹、CHAID決策樹、倒傳遞類神經建構企業財務之危機預警模型準確度及影響公司危機之重要關鍵因素。
實驗結果發現C5.0決策樹模型整體危機前三年準確度高於其他三者，其次為支持向量機模型。經濟危機之情況之下，本研究四種資料探勘方法建置之危機預警模型影響公司重要關鍵因素，「負債占資產比率」、「流動比率」、「稅前純益占實收資本比率」、「利息保障倍數比率」、「資產報酬率」、「每股盈餘」、「營業利益占實收資本比率」、「現金再投資比率」8個財務比率，提供企業與投資者做為參考預測的依據。

關鍵字

資料探勘 ； 分類法 ； 企業財務危機

並列摘要

The problem of Enterprise Financial Crisis has been the critical topic that financial researchers make their efforts to explore. In our study, we use the financial data of companies, which are collected from the New Taiwan Economic News database from May 2003 to April 2013 that conform to the definition of financial companies. Our study used a 1: 2 pairing sampling proportion as the ways of selecting samples. Thus our sampling financial data are collected from 52 companies that are under financial crisis and a group of controlling 104 companies that are claimed to be normal. The incorporating financial variables also include 19 financial variables suggested by Altman (1968) and Ohlson (1980). The analyzing methods of our experiment also include support vector machine, C5.0 decision tree, CHAID tree, and the back-propagation neural Network. We have construct an enterprise financial crisis early warning model and try to find the important key factors of the company that might cause the financial crisis impact.
At last, our study has found that the accuracy of overall three-year crisis calculated by C5.0 decision tree model is better than the other three methods. SVM model gets the second accuracy rate. Under the economic crisis and the importance of the key factors affecting the company in this crisis early warning model of Four build data mining method, the eight financial ratios: "Debts ratio", "Current ratio", "Per-tax income to capital", "Interest guarantee ", "Return on total assets", "Earning per share", "Operation income to capital", "Cash Re-investment Ratio," may provide the enterprises and investors as reference prediction basis.

並列關鍵字

Data Mining ； Classification ； Enterprise Financial Crisis

參考文獻

英文文獻

1.AI Magazine(1996) From Data Mining to Knowledge Discovery in Databases, Providence, Rhode Island July 27–31, 1997

2.Altman, E. I.(1968). Financial Ratios, Discriminate Analysis and The Prediction of Corporate Bankruptcy. Journal of Finance. 23, 589-609.

3.Beaver, W. (1966), Financial Ratios as Predictors of Failure, Journal of Accounting Research, pp.72－102.

4.Berry, J. A. & Linoff, G. (1997). Data Mining Techniques: For Marketing, Sales, and Customer Support. New York: John Wiley & Sons, Inc.