Research On Stock Selection Model Based On Investors' Attention And Social Networks

With the development of the Internet,the Internet has become an important part of people's lives.On the one hand,people obtain information they care about through the Internet and make their own decisions.On the other hand,the Internet has become an important way for people to express their attitudes and spread their opinions.Correspondingly,search engines and social networks are typical Internet products based on people's two kinds of demand.With big data era,people's behaviors are all recorded and can be used for analysis under the premise of protecting personal privacy.Therefore,behavioral finance researchers have conducted fruitful research using people's Internet behavior.Among them,the study of the limited attention of investors through the Internet search behavior and the use of public opinion platforms to reflect investor sentiment to predict the changes in the stock market has become a research hotspot in behavioral finance.It is on these foundations that the reptiles first obtain the stock-related Baidu index and the investor network commentary data represented by Sina shares,as well as the stock historical transaction-related data.For the structured processing of unstructured data,the sentiment classification of the textual sentiment corpus is built on the text data;the Baidu index is processed according to the previous method of paying attention to investors;finally,a deep learning network based on LSTM is built to predict the stock market.Finally,using the industrial neutral strategy and longshort strategy to filter the model prediction results constitute a portfolio backtesting.The research shows that the LSTM-based deep learning model can predict the stock price fluctuations.Adding the pre-processed investor sentiment data and the anomaly search volume calculated using Baidu index to the model can improve the model prediction results to some extent.The industry neutrality and volatility strategy can reduce the strategic risk while keeping the cumulative yield basically unchanged,and enhance the robustness of the trading strategy.Compared with the ShanghaiShenzhen 300 index,the trading strategy proposed in this study can obtain better returns and can control the risks within a more reasonable range.This study has made a certain degree of contribution to the deep-learning method in the stock market application.It combines traditional research content methods and artificial intelligence to provide new ideas for research in this field.In addition,it also has decisions on individual investors' stock investment.Certain reference value.