結合特徵選取與分群法和分類法建構財務危機預警模型

Combining feature selection,classification and clustering to construct models of Bankruptcy Prediction

謝易修(Yi-Hsiu Hsieh)

指導教授 ： 李維平

中原大學/商學院/資訊管理研究所/碩士(2014年)

https://doi.org/10.6840/cycu201400944

摘要

公司是否會發生財務危機對於投資者以及金融市場是一個重要的議題，而財務預警模型可以用來預測公司是否會發生財務危機。
先前學者的論文中，財務預警模型大多都只探討兩種方式，第一種是特徵選取與分類，這個方式的優點是可以提升模型的準確率和減少無效或不相關的特徵並增加效率。第二種是分群與分類，這個方式也可以增加模型的準確率與模型的穩定性。而本文不同的地方是嘗試結合上述兩種方式的優點來建構預警模型，並進行模型的比較。本文中特徵選取使用了逐步回歸分析、線性回歸和分類與迴歸樹來篩選變數，分群方法使用K平均演算法與自組織映射圖網路用來進行分群，而分類方法使用C4.5決策樹、倒傳遞類神經、最近鄰居法與隨機森林。資料使用國外期刊中較常使用的Australian、German等二個資料集，使用上述方式所組合的模型，進行準確率與型一錯誤率以及型二錯誤率的比較，找出最佳的預警模型。

關鍵字

財務危機 ； 特徵選取 ； 分群 ； 分類

並列摘要

Bankruptcy Prediction is the important issue for investors and financial markets, and the bankruptcy prediction model can be used to predict bankruptcy.
In previous papers, mostly only discuss bankruptcy prediction model in two methods, first is combining feature selection and classification, the advantages of this methods is the model can improve the accuracy and reduce ineffective or irrelevant features and increase efficiency. The second is combining clustering and classification, this method can also increase stability and accuracy of the model. And this article is try to combined two methods advantages to construct the bankruptcy prediction models and compare models. In this article Feature Selection methods use the stepwise regression analysis, linear regression and classification and regression trees to filter variables, clustering methods use k-means and self organizing map. Classification methods using C4.5 decision tree, backpropagation neural network, k-nearest neighbors algorithm and random forests. In this article use foreign journals more commonly used datasets Australian, German two datasets, combined two methods to construct bankruptcy prediction model, to compare the accuracy, type I type II error rates and find the best bankruptcy prediction model.

並列關鍵字

Bankruptcy ； Feature selection ； Classification ； Clustering

參考文獻

6.Beaver, W. H., (1966). Financial Ratios as Predictors of Failure. Journal of Accounting Research , 4, 71-102.

9.Blum, M., (1974). Failure Company Discriminant Analysis. Journal of Accounting Research ,12, 1-25.

10.Cover, T. & P. Hart. (1967). Nearest neighbor pattern classification, IEEE Transactions on Information Theory, 13, 21-27.

11.Elena Fedorova a, Evgenii Gilenko b,& Sergey Dovzhenko. (2013). Bankruptcy prediction for Russian companies: Application of combined classifiers, Expert Systems with Applications, 40, 7285–7293.

12.Etemadi, H., Rostamy, A. A. A. & Dehkordi, H. F. (2009). A genetic programming model for bankruptcy prediction: Empirical evidence from Iran. Expert Systems with Applications, 36, 3199-3207.