A Research On Stock Market Prediction Based On Text Analysis

The information age has come.More and more information can be directly accessed through the Internet,so is financial information.Meanwhile,through online financial community,investors can also fully communicate with each other about their opinions of financial market.So these data such as online review on relative website can reflect investors' emotional tendency towards financial market.Therefore,it is important to study the relationship between online review of investors and performance of financial market.Based on online stock views,this paper studied the influence of investors'emotional tendency on stock market.In order to explain the relationship,first,we introduced behavioral finance theory and text analysis technique which would be used in further research;and then,we compared several big online financial community based on website evaluation indexes and selected a appropriate online community as data source;next,we collected data from selected website by programming web crawler and pre-treat the raw date;we also did a empirical research based on posting situation and found that the online community was actually quite active which provided reliability to the research results;then we developed classification system based on SVM and classify the review data;in the final,we discussed the influence of online stock review on stock market in three aspects which was Shanghai Composite Index,certain industry stocks and individual stock,and established regression model to forecast the stock price.The result shows that there is a certain relationship between the returns of Shanghai Composite Index and unusual daily postings on online financial community and the Weekend Effect not only exists in stock market but also applies to the change of investing emotion in community.It also shows that there is a mutual influence between investing emotion and market performance.Then,based on the emotion information we collected from online reviews,we combined them with stock financial indexes and established a regression model to forecast the stock price,the fitting effect is actually pretty good.