网络投资者情绪与股票市场价格关系研究——基于文本挖掘技术分析

The Research on the Relationship between Network Investor Emotion and Stock Market Price——Empirical Analysis Based on Text Mining Technology

本文将东方财富网股吧的评论数据作为实证对象，利用文本挖掘技术构建金融领域的情感词典，通过贝叶斯方法将其合成网络情绪指数，应用ARMA-GARCH族模型分别刻画网络情绪与个股收益序列。结果表明：AKMA-GARCH族模型能有效解释网络情绪与股票收益的自相关性与异方差性；在短期内网络情绪对大多数个股的收益具有一定的预测作用，而个股收益对网络情绪的影响则具有较长时滞。

Behavioral finance points out that investor sentiment plays an important role in making investment decisions. The study of the impact of investor sentiment reflected in online media on the stock market has a strong significance. This paper takes the comment data of the Guba East-money as empirical object, uses the text mining technology to constructs a sentiment dictionary for financial field, classifies the bullishness and bearishness of sentiment contained in the single stock comment by Bayesian method, and computes the network investor sentiment index. Based on this, the ARMA-GARCH model is used to characterize the network emotion and individual stock return respectively. The results show that ARMA-GARCH model can effectively explain the autocorrelation and heteroscedasticity of the network sentiment time series and stock retums time series. And the results of Granger causality test of residuals show that network sentiment has a predictor of returns for most stocks in short period,while stocks＇ returns have a long time lag on the impact of network sentiment.