Web Search,Investor Sentiment And Stock Market

As big data and stock reform is becoming more important in our economic development in recent years, exploring the acting mechanism of big data information on the stock market has a very important practical significance. It has been proved by domestic and external research that, investor sentiment index can be used to predict stock market’s future trends and has a spillover effects on it, but it still has research gaps in using big data web search engine information to construction leading sentiment index. Shanghai Stock Exchange and Shenzhen Stock Exchange are the only two stock exchange in our country, which are also the main market of capital flows. Therefore, to build up the Investor Sentiment Index using web search information and to explore the mutual spillovers effect mechanism between Investor Sentiment Index and Stock Market Index are meaningful.This paper will focus on a multidimensional research on China’s Shanghai and Shenzhen stock market and Investor Sentiment Index, with the combination of the theory of statistical methods and time series models, and domestic and foreign research results or practical experience. Based on the understanding of the meaning of Investor Sentiment Index and Stock Market, this paper will constructs a "Stock Market" Google search words thesaurus by the method of text mining technology, and the extract the core information among the initial web search data using the time difference correlation coefficient method, random forest, CART, neural networks algorithm, and so on, and compute the Hu-Investor leading Index and Shen-Investor leading Index with the combination of PCA and RF methodologies. And then, we use VAR(n)-BEKK(1,1)-GARCH model to explore the spillover effects between Hu-Investor leading Index and00001, Shen-Investor leading Index and399001, Hu-Investor leading Index and Shen-Investor leading Index, and make some conclusion at last. The main conclusions are: First, random forests algorithm do much better in variables selection fields than other algorithms. Second, social text messages and web data have a good reference value to infer investor sentiment. Third, Investor Sentiment Index constructed by Google search data is a good future trend predictor. Fourth, Investor Sentiment Index shows different features in different stock market or in short-long term. Fifth, there is a linkage mechanism between Hu or Shen Investor Sentiment and its Stock market Index. Sixth, Investor Sentiment Indexes and stock market indexes are proved to be strong anti-self effect. Seventh, there are mutual spillover effects between Investor Index and stock market index both in Hu and Shen. Lastly, Investor Sentiment has across-spillover effects.