Font Size: a A A

Research On The Correlation Between Investor Sentiment And Stock Market Based On Text Mining

Posted on:2019-09-02Degree:MasterType:Thesis
Country:ChinaCandidate:L M YinFull Text:PDF
GTID:2429330566989774Subject:Management Science and Engineering
Abstract/Summary:PDF Full Text Request
Investors in the financial sector usually post stock comments on BBS,blogs,shares,weibo and other online information platforms to express their emotions and opinions.However,due to the systematic deviation of investors' irrational psychology and the uncertainty of their behavior,the number of comments will be large,the data is jumpy,the identification is low,and the quality is difficult to guarantee,these make the review information difficult to be directly used for the analysis of investor sentiment.But investor sentiment is an important factor of affecting stock market movements,therefore by text mining technology researching stock review information to reveal the effect of investor sentiment in the stock market has caused extensive concern of the industry and academia.Investor sentiment will have an important influence on the stock market trend.The network community contains abundant textual information about stock comments,and the investor sentiment analysis based on text mining provides support for relevant research.First,the crawler program is used to capture the stock comment data of the Shanghai composite index.Through the data cleaning,it clears the data that is irrelevant to the stock and the repeated stock comment data.Then,using the machine learning method and the semantic understanding of the emotion dictionary matching method,the stock comments of Shanghai stock index are classified and their effectiveness is compared.Then,taking into account the emotional tendency of the stock comment publishers and followers,the author constructs the "publisher-follower" emotional value SV of each comment,and builds the "publisher-follower" BSI investor sentiment indicator based on the form of logarithmic adjustment and additive product synthesis.Finally,the Composite Index of SSE Composite Index(CIt)and Returns(Rt)are respectively used as explanatory variables to establish the multiple regression model for empirical test.The study finds that(1)the sentiment classification method comparison experimental results show that for the overall market stock review data used in this paper,the semantic analysis method of emotional dictionary matching is better than the SVM machine learning method.(2)The empirical analysis results show that the "publisher-follower" BSI index constructed in this paper is significantly correlated with the price and yield of the Shanghai Composite Index,and that the BSI's ability to forecast the return rate is greater than the ability to predict the stock price and the degree of interpretation is more than 28%.(3)We also found that the market performance was positively correlated with the BSIt of the day and negatively correlated with the previous day's BSIt-1,indicating that the investor sentiment has a lagging effect: under the common driving of investor sentiment on the two days,the current sentiment towards the current period Market performance has the same direction,while the previous emotions have a correction effect on current market performance.
Keywords/Search Tags:stock comment, text mining, affective tendency, investor sentiment
PDF Full Text Request
Related items