Modeling And Evaluating Of Composite Stock Price Rise And Fall Prediction By Integrating Network Stock Comments And Public Opinions

In today's society,stock has become an important financial instrument widely accepted by the public.As the "Barometer" of economic operation,stock investment has also become the financial means of more and more people.The prediction of stock price plays an important role in realizing financial goals and stabilizing economic development.The rapid development of network technology at the same time.The Internet from a simple information release technology platform evolved into the main carrier of social media.And the impact of network information on the stock market is also increasingly prominent.More and more listed companies have opened official Weibo to release news and interact with shareholders.Investors are increasingly relying on online information to make investment decisions.They communicate and discuss in stock BBS,release information or comment on others'information,and form interactive exchanges with listed companies and other investors.In this context,it is of great academic and practical significance to add network information into the research of the stock industry and realize more accurate prediction of stock price and rise and fall.The main goal of this paper is to predict the rise and fall of stock prices by using the fundamental data of stock trading and the comments on the Internet.Construct a comprehensive prediction system with popularization and reference value.Realize mutual promotion between Internet public opinion analysis and stock finance research.Network crawler,natural language processing,machine learning,deep learning,econometric and other technologies are adopted in the research process,which has a strong interdisciplinary nature.First of all,this paper uses the web crawler technology to get the expectation of stock price fluctuation,the comment text and the research report of securities institutions on the stock industry in BBS,the Oriental fortune online stock bar.These network text segmentation processing,description statistics to achieve the initial processing and screening.At the same time,the trading data of the stocks in the csi 300 index in 2018 were obtained from the stock database.Secondly,natural language processing(NLP)is used to process the critical text.Analyze the emotional attitude behind the text and get the emotional score of the comment text.After weighting,matching and attenuation,the emotional scores of comments are corresponding to stocks and trading days.Then,the latest CatBoost algorithm suitable for binary prediction and LSTM algorithm good at realizing time series and sequence correlation are selected from the widely used and excellent machine learning algorithms to model the sorted data.Train CatBoost algorithm for 300 stocks,CatBoost algorithm for each stock,LSTM algorithm for each stock price.This paper studies the influencing factors of stock price rise and fall from different angles and the degree of influence to realize the preliminary prediction of stock price rise and fall.Finally,the stock price prediction results realized by the three algorithms are integrated with the Logit model.Considering the accuracy of each of the three algorithms in predicting the rise and fall of stock prices,a comprehensive prediction system for the rise and fall of stock prices is constructed to further improve the prediction accuracy and enhance the practical value of the model.The results show that online stock comments have a significant effect on the rise and fall of stock prices,and this effect has a lag effect.In addition,the stock's basic information and trading volume data on the rise and fall of the stock price is also obvious.The accuracy of CatBoost algorithm to predict the rise and fall of 300 stocks is 54.13%,and there is an obvious bias of "bearish" prediction.After the CatBoost algorithm training for each stock,the average accuracy of the prediction of stock price rise and fall increased to 60.67%,and the deviation of the prediction was also significantly alleviated.Then establish LSTM neural network for the stock price of each stock,the average accuracy of the prediction of stock price rise and fall further increased to 64.22%.After the above three algorithms compound the forecast results of stock price rise and fall by Logit model,the average accuracy reaches 70.20%.Meanwhile,the measured accuracy of stock price fall is very high,which can effectively help investors identify risks.The conclusions obtained in this paper are the average results of the constituent stocks of CSI 300 index,which effectively eliminates the randomness of high or low accuracy caused by different stock selection,and has stronger promotion value.To sum up,this paper uses abundant network comment text and stock trading data to train emerging machine learning algorithms.Design the comprehensive model of stock price rise and fall.It not only proves that the network stock comment has a significant influence on the stock price rise and fall,but also obtains a good effect on the prediction of the stock price rise and fall of Shanghai and Shenzhen 300 constituent stocks.It has certain popularization value in academic field and practical field.