許多研究著重於通過使用財務比率或將其與公司治理指標相結合來預測破產，而本研究追求的是使用公司負面新聞事件來預測公司的第一次危機，而這個危機會對公司帶來實際上的損害，這可能是防止他們破產的第一步。 在看門狗資料庫中所記錄每家上市上櫃公司的事件資訊，可用於預測公司的財務危機。 本研究採用統計學和機器學習方法用危機前三年的事件作為特徵來預測公司發生的第一次危機，此方法受研究問題的推動：實驗所選擇的自三年前開始的看門狗事件是否有助於提高預測第一次的危機表現？ 為了瞭解這個問題，我們提出了一個模型是通過財務比率和看門狗事件中的五種事件的組合所建立的。;While many studies focused on predicting bankruptcy by using financial ratios or combining them with corporate governance indicators, this study pursue a crisis events which causes lose to a company to predict their first crisis which can be an initial step to prevent them to be bankrupt in the future. Recorded events information-gathered by watchdog dataset of Taiwan listed companies-before the financial crisis happened in a company can be used to predict financial crisis in a company. This study focus on the prediction start from three year periods before first crisis happened in a company using statistical and machine learning method which is motivated with the research question: could the chosen watchdog events help to improve the performance of crisis prediction start from three years before first crisis? To acknowledge this question, a proposed model is built by a combination of financial ratios and five events gathered by watchdog using Taiwan listed companies.
