投资者情绪与证券投资收益:机器学习对网络社交媒体的文本分析  

Stock Returns and Investor Sentiment:Machine Learning for Textual Analysis of Online Social Media

作　　者：万漪萍 赵岩 赵留彦[2] WAN Yiping;ZHAO Yan;ZHAN Liuyan(Peking University)

机构地区：[1]北京大学软件与微电子学院金融信息工程系 [2]北京大学经济学院

出　　处：《金融市场研究》2024年第5期65-77,共13页Financial Market Research

摘　　要：本文研究投资者有关个股的情绪对股票收益的预测价值。我们通过网络爬虫收集了线上投资者论坛中有关上证50指数成分股在2019—2023年的700多万条评论信息,通过自然语言处理进行情感分类,从而构建了关于每只股票的周度投资者情绪指数。研究发现,投资者看涨情绪指数和关注度指数对于短期股价变化具有显著的预测能力,并且基于这两个因子所构建的量化投资策略能够获得明显超额收益。因而,在考察投资者情绪的效应时,研究者不仅要考虑到整体市场情绪,还应该关注投资者有关个股情绪的异质性。这些发现提供了有效市场理论的反面证据。This paper assesses the link between investor sentiment and stock investment returns.Researchers made use of a web crawler to capture 7 million posts on the constituent stocks of the Shanghai Stock Exchange 50 Index by means of the online Eastmoney Investor Forum from 2019 to 2023.Relying on textual analysis of the captured social media posts,a weekly measure of investor sentiment was created for each stock.The study found that the weekly measures of bullish investor sentiment and investor attention had significant predictive ability for short-term stock returns,and a quantitative investment strategy based on these measures could obtain substantial excess returns.These findings were consistent across a number of different models and specifications,providing further evidence against the efficient market hypothesis.Moreover,while market sentiment plays a role in explaining the overall mispricing of assets,stock-specific sentiment is likely to be informative.

关 键 词：投资者情绪 投资者关注度 社交媒体 文本分析 

分 类 号：F83[经济管理—金融学]