運用文字探勘與機器學習技術探討COVID-19疫情趨緩新聞對企業異常報酬之研究

A Study on the Impact of COVID-19 Recovery News on Abnormal Returns of Firms Using Text Mining and Machine Learning Techniques

近年來各個產業無不受到COVID-19疫情的影響，而當疫情有變化時，對於股市漲跌也會有相關的影響。對於投資人而言，新聞媒體的訊息是最容易取得的管道之一，而其中有些報導會帶有一些情緒字眼，這些字眼容易造成投資人預期心理的影響，進而導致股價的漲跌。
本研究透過對疫情趨緩新聞的篩選，運用文字探勘來萃取疫情趨緩關鍵字，再利用決策樹、隨機森林與類神經網路工具來找到疫情趨緩關鍵字與企業股價異常報酬之間的關聯性，最後本研究得出結論，疫情趨緩新聞會影響股市產生異常報酬並多為短期，且內文出現關鍵詞的次數越高，越容易產生正異常報酬。

In recent years, various industries have been significantly affected by the COVID-19 pandemic, and fluctuations in the pandemic's status have had corresponding impacts on the stock market. For investors, news media is one of the most accessible sources of information. Some news reports contain emotionally charged language, which can influence investor expectations and subsequently lead to changes in stock prices.
This study focuses on analyzing news related to the easing of the COVID-19 pandemic. Using text mining techniques, we extract keywords indicative of pandemic mitigation. We then apply decision trees, random forests, and neural networks to explore the relationship between these keywords and abnormal stock returns. The findings suggest that news concerning the easing of the pandemic can indeed generate abnormal returns in the stock market, typically in the short term. Furthermore, a higher frequency of mitigation-related keywords in the news content is more likely to be associated with positive abnormal returns.