企業危機預警模式之研究－－DEA-DA邏輯斯迴歸與類神經網路之應用

A study on Prediction Models of Financial Distress —A comparison of DEA-DA、N

為有效建立企業之危機預警模式，本研究利用過去相關文獻常用之財務指標（包含應收帳款週轉率、流動比率、每股盈餘與負債比率），以DEA-DA、邏輯斯迴歸與類神經網路等三種不同之研究方法，來比較其判別正確率與預測正確率。研究結果發現三種方法所建立的預測模式，其判別正確率與預測正確率均達80%以上，其中邏輯斯迴歸判別之敏感度僅為33.3%，但精確度為100%，而DEA-DA及類神經網路之敏感度均達60%以上，精確度均超過85%。 In this paper, we first survey the relative literatures to find the financialindicators. And then we use the methods of DEA-DA、neural network andlogistic regression by the financial indicators to establish the prediction models offinancial distress. In the main results, we discover that the hit rates of these threemodels are at least 80%. Furthermore, the sensitivity is 33.3% and thespecificity is 100% for the logistic regression model. The sensitivity is more than60% and the specificity are more than 60% and the specificity are more than 85%for the DEA-DA and neural network models.