以往的研究廣泛使用財務比率和市場變數來構建財務困境預測模型。 Altman ‘s Z-Score在學術研究中已成為預測破產最常用的財務比率。 可以根據公司在市場上的一些變數（公司規模、市值、總回報、回報波動性、貝塔係數、違約距離和違約概率）預測公司現有的財務困境。 也有一些研究常見著名的機器學習算灋預測破產，但是之前沒有任何研究試圖將這七個基於市場的變數作為綜合市場變數，與集成學習方法相結合去預測破產的Altman分數。 以前的研究認為，stacking ensemble learning方法比單個分類器更能提高預測結果。 本研究旨在使用stacking ensemble learning方法結合邏輯回歸於5個Altman變數和7個基於市場的綜合變數作為預測因數，預測臺灣公司的財務困境。 現時的結果表明，與僅使用5個Altman變數相比，將7個基於市場的綜合變數添加到5個Altman變數中可以提高財務困境預測的效能。 此外，本研究還發現7個綜合市場變數和5個Altman變數之間存在著很强的聯系。 本研究的結果將解釋哪些基於市場的綜合變數會影響預測結果。 這些發現還將通過提供有關財務比率和市場變數的資訊，幫助公共投資做出貸款決策。 ;Previous research had extensively used financial ratios and market variables to build their model of financial distress Prediction. Altman′s Z-Score has become the most commonly used financial ratio for identify bankruptcy, particularly in academic studies. Firm size, market cap, total return, return volatility, beta, distance to default, and probability to default are market-based variables that can forecast financial distress. The previous study also predicted bankruptcy using several well-known machine learning algorithms. However, no previous research attempted to combine these seven market-based variables as comprehensive market variables with Altman variables using the stacking ensemble learning method to predict bankruptcy. Prior studies believe that the stacking ensemble learning method can improve the Prediction result more than a single classifier. This study aims to use the stacking ensemble learning method in conjunction with Logistic Regression to predict financial distress in Taiwanese companies by combining 5 Altman variables and 7 comprehensive market-based variables as predictors. The result shows that adding the 7 comprehensive market-based variables to 5 Altman variables improves the performance of financial distress Prediction than only using 5 Altman variables. Also, this research found a strong connection between 7 comprehensive market-based variables and 5 Altman variables. The findings of this study will explain which comprehensive market-based variables influence the Prediction outcome. These findings will also aid public investment in lending decisions by providing information on financial ratios and market variables.
