Limited Attention And A Share Market Price Regression Forcast

With the innovation of current Internet technology and the rapid development of social media,people can express their opinions online in real time,which has become a new way of communication.Online comments have become a way for many people to express their views.For the securities market,a lot of professional stock BBS has been developed on the Internet for people to express their views on stock investment.In-depth analysis and mining of information released by investors on the network platform is not only our way to understand investor behavior and emotions,but also our basic means of mining investor emotions.In recent years,investor sentiment has become one of the more concerned indicators in stock investment.From the early structured indicators to text mining,it has attracted the attention of more and more financial enterprises and regulators,and has gradually become one of the hot issues in the research of scholars.This paper focuses on the massive text data on the network platform,extracts and analyzes the unstructured text data through text mining,compares the information of domestic stock BBS platform,and selects xueqiu net BBS as the basis of this paper to study the limited attention of investors.BBS is an open stock exchange BBS.Online comments and other information in BBS can better reflect investors' instant thoughts on relevant stocks from one side.In recent years,domestic research on mining and stock forecasting has made some achievements.Based on the previous research results,this paper mainly studies the relationship between investors' limited attention and stock prediction,and compares the different methods of machine learningIn terms of theoretical research,this paper first defines limited attention and different theoretical bases,and then puts forward the research significance of constructing investors'limited attention index based on stock online comment data based on text mining on the basis of existing research results.From the perspective of emotional analysis,all text information is classified into three categories:positive text,negative text and neutral text through natural language processing.Furthermore,after classifying and processing the information,a vector space model and an emotional feature model were established.Support vector machine regression and logistic regression were used to quantify the text classification results.In terms of practical application research,this paper firstly takes the text data of sse 50 index online stock comments obtained from BBS of snowball network as the research object,expounds the process of data acquisition and preprocessing,and further constructs the investors'limited attention index based on text mining.At the same time,this paper also selects the index of structured data as the independent variable together with the index of limited attention of investors,and selects the closing price of sse 50 index as the dependent variable.Next,this paper studied the correlation between the corresponding independent variables and dependent variables,screened out the corresponding indicators related to the dependent variables through stationarity test and granger causality test,and then constructed two models for the corresponding indicators to test the accuracy of the independent variable indicators under different models in the prediction of the stock market.The research shows that there is a correlation between bullish sentiment index and general sentiment index and stock market closing price,and investors' limited attention index based on support vector machine can predict the stock market more accurately.