文本情緒的報酬可預測性： 以中國股市?例;The Return Predictability of Text-based Sentiment: Evidence from China

NCUIR > School of Management at National Central University > Graduate Institute of Finance > Electronic Thesis & Dissertation > Item 987654321/87141

Please use this identifier to cite or link to this item: http://ir.lib.ncu.edu.tw/handle/987654321/87141

Title:	文本情緒的報酬可預測性：以中國股市?例;The Return Predictability of Text-based Sentiment: Evidence from China
Authors:	林奇斌;Lin, Qi-Bin
Contributors:	財務金融學系
Keywords:	投資者情緒;文本探勘;XgBoost模型;報酬預測;investor sentiment;text-mining;XgBoost model;return predictability
Date:	2021-10-25
Issue Date:	2021-12-07 15:07:16 (UTC+8)
Publisher:	國立中央大學
Abstract:	本研究以2019年6月至2019年12月中國大陸上證A股市場上市公司為樣本，爬取東方財富網站旗下論壇「股吧」個人用戶發帖內容，利用XGBoost模型對文本情緒進行提取並構建投資者情緒指標，探討投資者情緒對於大陸個股股票報酬之影響。實證結果發現情緒指標能夠預測下一期股票報酬，隨後股票報酬發生反轉；對於小市值公司而言，投資者情緒對下一期個股股票報酬有較?的影響，但隨後的反轉更加?烈；投資者偏好在股票交易時段進行發貼，且該時段帖子字數普遍較短，非交易時段情緒指標對於股票報酬有更?的預測能力，反轉較小。;This paper takes the Shanghai A-share stock as sample, uses python to crawl data from the post content of individual users of the forum "Guba" of eastmoney website, and uses the XGBoost model to extract the sentiment of the text to construct investor sentiment indicators. This study explores how investor sentiment affects the individual stock returns in China. The results show that investor sentiment can predict the next period individual stock return, yet then the stock return exhibits reversal. The return of small stocks are more affected by investor sentiment. Individual investors prefer to send messages during the stock trading hours, and posts during this period are generally short in words. During non-trading hours investor sentiment has stronger predictive power for stock returns, with smaller reversals.
Appears in Collections:	[Graduate Institute of Finance] Electronic Thesis & Dissertation

Files in This Item:

File	Description	Size	Format
index.html		0Kb	HTML	167	View/Open

社群 sharing

Loading...