›› 2018, Vol. 30 ›› Issue (7): 16-25.

• 经济与金融管理 • 上一篇    下一篇

基于股吧信息的投资者情绪与极端收益的可预测性研究

金秀, 姜尚伟, 苑莹   

  1. 东北大学工商管理学院, 沈阳 110169
  • 收稿日期:2016-04-25 出版日期:2018-07-28 发布日期:2018-07-21
  • 作者简介:金秀,东北大学工商管理学院教授,博士生导师;姜尚伟,东北大学工商管理学院博士研究生;苑莹,东北大学工商管理学院教授,博士生导师。
  • 基金资助:

    国家自然科学基金项目(71473033;71571041)。

Investor Sentiment from Guba Messages and the Predictability of Stock Extreme Returns

Jin Xiu, Jiang Shangwei, Yuan Ying   

  1. College of Business Administration, Northeastern University, Shenyang 110169
  • Received:2016-04-25 Online:2018-07-28 Published:2018-07-21

摘要:

随着互联网的广泛普及,基于互联网平台的投资者情绪对股市的影响研究,为情绪与股市关系研究注入了新的活力。本文首次采用Bayes分类算法对股吧信息分类,从基于质化信息的"情绪基调"、基于量化信息的"张贴程度"和基于强度信息的"关注水平"三个维度构建投资者情绪指数,并从极端收益视角深入研究投资者情绪与上证指数的关系。研究发现,基于Bayes分类算法的投资者情绪指数,在解释上证指数变动趋势上具有优势;投资者情绪对不同趋势极端收益的影响存在非对称性,对下跌趋势极端收益有显著可预测性。研究结论能够为投资者投资和监管者完善市场建设提供决策依据。

关键词: 股吧信息, Bayes分类算法, 投资者情绪, 极端收益

Abstract:

With the widespread popularity of Internet, the impact of investor sentiment based on the Internet platform on the stock market has injected new vitality into the relationship between the sentiment and the stock market. This paper first classifies guba (online stock forum) messages by using the Bayes classification algorithm, develops a comprehensive index to measure investor sentiment from three dimensions:"tone" based on the quality information, "exposure" based on quantity information and "attention" based on the intensity information. On that basis, this paper makes a research deep into the relationship between investor sentiment and extreme returns. The results show that investor sentiment index based on Bayes classification algorithm has more advantages in explaining the change trend of Shanghai index. Further evidence indicates that the effects of investor sentiment on the different trend of extreme returns are asymmetric. Specifically, there is a significant predictability for the positive extreme returns. The conclusion can provide advices for investors to invest and for regulators to optimize the market construction.

Key words: guba messages, Bayes classification algorithm, investor sentiment, extreme returns