以文字探勘分析社群媒體情緒對於股市報酬率之影響：以臺灣中大型股為例

Abstract

資訊科技日新月異，其建立了訊息管道的便利性和即時性，亦成就了訊息爆炸與「害怕錯過」的社群焦慮。以投資角度而言，「羊群效應」來自於投資者的有限理性，形成從眾心理，藉以社群平台的普及性與同質群聚效應，投資者對社群平台漸成依賴，隨時可於社群平台發佈及獲取相關資訊，並對其取得之資訊產生即時反應行為或效果。本研究係探討社群媒體情緒對股市報酬率之影響，檢視 2019 年 1 月至 2022 年 2 月期間，以台灣中大型股以及台灣加權指數為樣本，透過文字探勘方式，量化社群平台中投資者發佈之文章與新聞內容，採以計算且結構化新聞及投資者情緒，並將情緒區分為正向及負向，藉此予以探討社群平台上之情緒對於投資者交易決策的影響。研究實證結果顯示，日平均新聞情緒及投資者情緒之遞延效果兩日後與報酬無顯著關係，而負向的新聞情緒及投資者情緒相對於正向的情緒而言，皆對於報酬率的影響較大；本研究探討之社群平台中，以 CMoney 平台對台灣股市正向影響最大；研究檢視期間，新聞情緒的預測能力較投資者情緒強；另嚴重特殊傳染性肺炎(COVID-19)疫情期間，大盤、個股之情緒預測能力都較疫情前強；電子類股報酬率受情緒之影響最大。

The rapid development of information technology has created the convenience andimmediacy of information channels, and has also contributed to the explosion of information and the social community anxiety of"fearing for missing out something." From an investment perspective, the "herd effect" comes from the limited rationality of investors to form the herd mentality. Investors gradually become dependent on social platforms with the popularity of social platforms and the homogeneous crowd gathering effect. They can post or obtain relevant information anytime, and generate immediate responsive actions or effects to the information they have obtained. This study examines the impact of social media sentiment on stock market returns. This study examines a sample of Taiwan mid-large cap stocks and Taiwan Capitalization Weighted Index (TAIEX) from January 2019 to February 2022. Then, the articles and news contents posted by investors on the social media platforms are quantified through text mining. The news and investor sentiments are calculated and structured, and the sentiments aredivided into positive or negative; by doing so, the impact of sentiments of the social platform on investors' trading decisions are investigated. The empirical results of the research show that the deferred effect of daily average news sentiment and investor sentiment has no significant relationship with remunerations after two days. While negative news sentiment and investor sentiment have a greater impact on the rate of return than positive sentiment; among the social media platforms discussed in this study, the CMoney platform has the greatest positive impact on the Taiwan stock market. The ability to predict news sentiment was stronger than that of investors during the period of research study. The sentiment predictability of the main stock market and individual stocks was stronger during the severe COVID-19 pandemic than before the epidemic. The return rates of electronics stocks are most affected by the sentiments.