| DC 欄位 |
值 |
語言 |
| DC.contributor | 會計研究所 | zh_TW |
| DC.creator | 謝佳芳 | zh_TW |
| DC.creator | Chia-Fang Hsieh | en_US |
| dc.date.accessioned | 2020-7-20T07:39:07Z | |
| dc.date.available | 2020-7-20T07:39:07Z | |
| dc.date.issued | 2020 | |
| dc.identifier.uri | http://ir.lib.ncu.edu.tw:88/thesis/view_etd.asp?URN=107430014 | |
| dc.contributor.department | 會計研究所 | zh_TW |
| DC.description | 國立中央大學 | zh_TW |
| DC.description | National Central University | en_US |
| dc.description.abstract | 資本市場瞬息萬變，企業為了永續經營，除了要維持彈性以因應變動，如何避免危機也是一個重要的課題，本研究以Altman (1968)中的Z-Score建立財務變數模型，再建立一個整合財務與資訊優勢者異常事件變數的模型，並將上述變數丟入羅吉斯迴歸、隨機森林與決策樹等三種不同的模型進行比較分析。
研究結果顯示財務變數累積盈餘能力與破產前下跌規模為危機預測中較重要的指標；股權結構部分，高管持股、董監持股與董監持股質押呈現正向顯著，顯示當高管或董監持股比率越低、董監持股質押比率越高，越容易發生危機，符合「利益收斂假說」；內部異動部分則發現高管異動於危機前二年為正向顯著，更進一步發現高管持股於危機前二年與前三年呈現負向顯著，顯示高管提早將手中持股出脫並於日後離職，因此看到高管持股與高管異動於危機前一年呈現不顯著的現象；外部監督部分會計師繼續經營假設疑慮與變更會計師事務所皆與危機事件呈現不顯著現象；外部資訊部分股價暴跌或異常則呈現危機前二年高度正向顯著，顯示投資人提前接收企業資訊後於資本市場反應了公司價值。
綜上所述可以發現，資訊優勢者異常事件補足了財務變數所缺失之部分，甚至可以提早至危機前三年預測未來危機事件，發現加入資訊優勢者異常事件變數後能提高模型對危機事件的模型整體預測力與模型配適度，綜合結果隨機森林居冠，羅吉斯迴歸次之，最後才是決策樹。 | zh_TW |
| dc.description.abstract | Crisis Forecasting have always been a discussion thourgh out the history. People have been searching for better variables to predict crisis events in a more precise way. In this study, we find out that there is a group of people can retrieve information before the capital market. That means they can do some strategies before outsiders, that’s the definition of information superior. This study shows that the higher the extent of shares as collateral by the board of directors and the lower the extent of shares holded by the managers or the board of directors, the possibilities of crisis go higher. In addition, the empirical results show that the abnormal downsize of stock prices appears before two years of crisis events as the decline of the extent of shares holded by the managers and the maneger turnover. That means the change of managers and the decline percentage of shares holding result in higher agency problems, which are signals of crisis events.
As a results, the supplement of events in information superiors make the accuracy, AUC and F1 Score in Logistic Regression, Random Forest and Decision Tree gets higher than pure Z-Score model. The variables we use have even greater ability to Crisis forecasting than pure Z-Score model in two or three years before crisis events. In the combined evaluation between models, Random Forest have better accuracy than Logistic Regression, with Decision Tree the lowest one. | en_US |
| DC.subject | 公司治理 | zh_TW |
| DC.subject | 危機預測 | zh_TW |
| DC.subject | 機器學習 | zh_TW |
| DC.subject | 資訊優勢者異常事件 | zh_TW |
| DC.subject | Corporate Governance | en_US |
| DC.subject | Crisis Forecasting | en_US |
| DC.subject | Machine Learning | en_US |
| DC.subject | Information Superior Abnormal Events | en_US |
| DC.title | 危機預測之探討–以財務與資訊優勢者異常事件為例 | zh_TW |
| dc.language.iso | zh-TW | zh-TW |
| DC.title | The Research of Crisis Forecasting–Take Financial and Information Superior Events for Examples | en_US |
| DC.type | 博碩士論文 | zh_TW |
| DC.type | thesis | en_US |
| DC.publisher | National Central University | en_US |