A Correlation Study Between Online Information And Stock Market

With the progress of economic development and information technology, the Internet has become the main channel for the Internet users' emotional communication and information dissemination. Nowadays, vast amounts of information generated every day on the Internet. Information is an important basis for investors to make investment decisions. Most Investors' investment failed because of information asymmetry. In this context, we explore the correlation mechanism of network information and security market which has important theoretical and practical significance. If we get understand of this correlation mechanism, we can obtain useful information efficiently, and we can extract the valuable knowledge for security market forecasting, which can provide a strong basis for investors'market analysis and decision making.Because the information from stock forum are most relevant to stock market, what this paper wants to explore is whether the information from the network forum will have an influence in security market, especially in stock yield, trading volume, volatility and other variables. Our approach is to use software to collect all posting text of the past year's 50 stocks of SSE 50 Index from Eastern wealth net. We get 2.78 million records totally. Then, we use Naive Bayesian classification algorithm to classify all these posts into "positive "" neutral "," negative "categories, and build investor's sentiment indicators and the investor's agreement based on the classification results.According to empirical data and the statistical description of the raw data, we propose three assumptions:(1) trading volume has a positive correlation with investors'posting volume; (2) the securities market yields have a positive correlation with investors' sentiment index; (3) stock's volatility has a negative correlation with investors' agreement. Then, we use SSE 50 Index as an example, we analysis the correlation between Internet Variables and Security Variables.According to the correlation analysis' result, we preliminary judge Hypotheses 1 and 2 is established, which Hypotheses 3 is not established.In order to explore the further relationship between internet variables and security variables, this paper uses 50 stocks' data to make panel data regression with both contain lagged variables and no lagged variables. The results of panel data regression just confirm our initial judgment described above. All in all, we have the following conclusions:(1) Investors' post time is relatively fixed. The first peak of posting volume appears at 10:00 am, the next is 14:00 pm. The third peak appears at around 21:00. (2) Network Information variables do have correlation with the stock market variables. Investor sentiment index and investor agreement may be used to explain stock market returns. If investor sentiment index rises, or the investor agreement is high, the securities yields may be high at the same period. In addition, the posting volume can explain securities trading volume well. (3) Internet Information variables can use to predict the stock market variables. By using panel data regression model which containing lagged variables, we find historical information from the network can be used to predict security variables including stock return and volume. But volatility can't be predicted.