隨著我國經濟的持續成長，我國的資本市場的發展亦漸趨完備，但卻始終有地雷股存在，上市公司發生危機，往往讓一般投資大眾的血汗錢成為泡影，近年來，我國政府強調公司監理的推動，並成立了金融監督管理委員會，期望可以籍由更完整的監控機制來預防公司惡意倒閉的成事發生。 本研究即為考量公司監理之重要性以及預測財務危機研究方法的多元性，決定採用多變量區別分析、Logit迴歸分析、類神經網絡等研究方法，結合公司監理方面相關變數，期望可以找出最適合我國上市公司的財務預警模式，本研究先以因素分析法來進行篩選變數，而後再帶入區別分析模式以及Logit迴歸分析，結果發現： l.流動性指標與償債性指標之解釋變異能力最佳，推測公司短期債務過高為發生財務危機的主因。 2.三種預測模式中，以類神經網路的預測效果最佳，危機前一年至危機前三年之分類正確率為93.75%78.12%78.12%。 3.本研究亦發現三種模式於測式樣本中之正確率均高於訓練樣本，本研究認為主因為政府推動之公司監理已初步達成效果。

In Taiwan, capital market is getting more mature; but sometimes there are still some listed companies in financial crisis which may cause significant loss of investors. Recently, Taiwan government sets up Financial Supervisory Commission which focuses the importance of corporate governance and monitor companies from financial crises such as fraudulent insolvency. The study adopts (a) Discriminant analysis, (b) Logit regression analysis and (c) Artificial Neural-Network methods to predict company financial crises. The purpose of this study is to find out better model to predict listed-company financial warning which may result in financial crisis. The findings of this study include: 1. The liquidity ratio and the coverage ratio are better variables to explain the variance of samples. It suggests that companies which have high short-term debts are the main reason to cause financial crises. 2. Neural-Network makes the most accurate prediction of those three methods. The accuracy of anticipation in past three years of financial crises is 93.75% 78.12% 78.12% respectively. 3. The study also reveals that the accuracy of anticipation of the testing samples is better than the training samples. We may recommend that the government enforcement on corporate governance has some effectiveness.