財務危機預測一直以來都是一個重要的主題，吸引了世界各地的投資者以及學者的關注，從早期的統計學方法到現在的機器學習演算法，如何利用財務指標與其它有用的指標參數建出有效且準確的模型，更是一個被廣為討論的問題。我們過去曾使用台灣資料集証明財務指標加上公司治理指標當特徵，可以得到比較好的預測結果；但是在美國資料集底下，卻沒有這樣的趨勢。因為在美國資料集我們的公司治理指標搜集不多，所以本研究提出一個使用財務指標加上公司治理指標的stacking ensemble演算法，並且証實在特定的cost ratios底下，使用本實驗提出的演算法是更好的。;Financial distress prediction (FDP) is an important topic. There are many investors and researchers focus on this question, From statistical methods earlier to machine learning algorithm today, how to use financial ratios and other potential feature to build a better model is a wildly studied question. We have proved that build model by corporate governance indicator and financial ratios can improve performance in Taiwan dataset, but not in USA data set. For the reason we cannot collect as much corporate governance indicators as Taiwan dataset, we proposed a stacking ensemble algorithm by using financial ratios and corporate governance indicators, and prove that in specific cost ratios, this algorithm is a better way in FDP.
