跨媒體平台事件與台灣股票市場之關聯性研究

股票市場為一個國家或地區重要的資本市場之一，影響股票市場的因素非常多，景氣與金融狀況、投資人的群眾心理等等都可能造成股票市場的震盪。近年來，隨著網路與科技的發達，網路逐漸成為現實世界的縮影，社群媒體成為人們獲取訊息及接觸大眾意見的重要來源之一，然而目前對於多媒體平台與股票市場的研究中，並未觀察社群媒體上發生的特殊事件；此外，大多數研究並未針對跨平台交互影響之事件對股市的影響進行研究。因此本研究欲找出這些多媒體平台中發生的事件，觀察事件與股票市場的關聯性，並觀察跨平台事件是否會比單一平台對股票市場的影響更為重要。本研究透過兩種事件偵測模型孤立森林 (Isolation Forest, iForest) 以及指數加權移動平均控制圖 (EWMA) 對 PTT、Wikipedia 以及兩大數位新聞媒體平台 (中時電子報、聯合新聞網) 進行事件偵測，找出與 207 間台灣上市公司相關的事件，並進一步進行情感分類，將事件區分為正、負向事件，最後再透過事件研究法 (Event Study) 檢測本研究偵測出之事件，對應於股票市場是否會產生顯著的異常報酬率 (abnormal returns)。實驗結果發現，利用事件偵測方法從多媒體平台中偵測到的事件的確與股票報酬率有所關聯；此外，本實驗也發現跨平台事件比起單一平台事件的確會對應到更高的股票異常報酬率結果，其中又以利用 Wikipedia 點擊量來搭配新聞或 PTT 進行事件偵測能夠發揮出最好的效果；而在加入情感後，除了能辨別出對股票市場的正負走向影響外，事件對應之平均異常報酬率又有顯著的上升；最後本實驗還發現透過找出正確的事件偵測組合，能夠從無法預測未來異常報酬率變成能夠預測到未來一天異常報酬率的變動，進而依此做出交易決策。

關鍵字

異常偵測；社群媒體；孤立森林；指數加權移動平均；事件研究法

並列摘要

A stock market is one of the most important capital markets in a country or an area. Many factors can affect stock markets, for example, the overall economic conditions or the popular mind of investors. As Internet and technology have developed faster and faster nowadays, Internet has become a miniature of real life. Also, social media has become one of the most critical sources of people getting information. However, in current studies on social media and stock markets, no research has observed the special events on social media, and few researches focus on the influences of cross-media events on stock markets. Therefore, this paper aims to find out cross-media special events, observe the relation between events and Taiwan stock market, and further, observe that whether cross-media events have greater impacts on stock market than single-platform events. This research detects events on PTT, Wikipedia and two digital news media platforms, China Times (中時電子報) and United Daily News (聯合新聞網), through two event detection models, Isolation Forest (iForest) and Exponential Weighted Moving Average (EWMA) Control Chart. Through this process, events about 207 Taiwan listed companies can be found. Events are then classified into positive and negative events through sentiment classification. Event Study is adapted to observe events detected from the research, and to further observe if these events correspond to significant abnormal returns in Taiwan stock market. This research finds that events on social media that are detected through two models are actually related to returns on Taiwan stock market. Furthermore, this research also finds that compared to single-platform events, cross-media events do correspond to higher abnormal returns. Among all platform combinations, the combination of Wikipedia traffic volume with news or PTT posts can have better detected results. Sentimental classification results can further classify the direction of stock market and get better abnormal returns. This research also finds that through the adapted process, future abnormal returns can be predicted if adapting correct platform combinations. Thus, investors can make trade decision according to the prediction results.

並列關鍵字

Anomaly detection ； Social media ； Isolation Forest ； EWMA ； Event Study

參考文獻

Alanyali, M., Moat, H. S., & Preis, T. (2013). Quantifying the relationship between financial news and the stock market. Scientific Reports, 3, 3578.

Andersen, T. G., Bollerslev, T., Diebold, F. X., & Ebens, H. (2001). The distribution of realized stock return volatility. Journal of Financial Economics, 61(1), 43-76.

Bollen, J., Mao, H., & Zeng, X. (2011). Twitter mood predicts the stock market. Journal of Computational Science, 2(1), 1-8.

Boudoukh, J., Feldman, R., Kogan, S., & Richardson, M. (2013). Which news moves stock prices? a textual analysis (No. w18725). National Bureau of Economic Research.

Chandola, V., Banerjee, A., & Kumar, V. (2009). Anomaly detection: a survey. ACM computing surveys (CSUR), 41(3), 15.

國際替代計量

跨媒體平台事件與台灣股票市場之關聯性研究

全文下載

主題瀏覽