本研究運用KMV模型和倒傳遞類神經網路模型進行信用評等,並與台灣經濟新報評等比較分類之正確性。爲了比較不同模型的判別能力,除一般財務變數外亦加入TCRI評分等級及違約間距等非財務變數。結果顯示,在信用評等的能力上,台灣經濟新報與倒傳遞類神經網路相近,KMV模型較差。另外,在公司財務危機的預測方面,我們亦透過CAP及ROC曲線檢驗模型效力,結果以C4.5決策樹模型最佳,類神經網路模型次之。
The study apply KMV model and Backpropagation Neural Networks (BPN) in credit rating. Furthermore, we compare the correct rate with Taiwan Corporation Credit Rating Index (TCRI) of The Taiwan Economic Journal Corporation (TEJ). For the comparison of classification, we also add TCRI index and Distance from Default (DD) as non-financial variables except financial variables. Empirical results show that the TCRI is near to ANN in the ability of credit rating, KMV model is worse. What is more, in the effectiveness of financial distress prediction, we further utilize CAP and ROC curves to test the performances of forecasting models. Results reveal that C4.5 decision tree model is the best and then is Backpropagation Neural Networks.