Empirical Research On Stock Price Forecasting Based On Deep Learning Model

In this paper,natural language processing and deep learning model are used in this paper to study the effect of public opinion information on the prediction of securities price prediction model.Research mainly used in natural language processing to build emotional tendency analysis dictionary method,study on the financial news and securities firms more accurate emotional information extraction and emotional judgement;The deep learning model selects the recursive neural network(RNN)as the basic model in this paper.In this paper,the content of the research is systematically elaborated through six chapters.The first chapter mainly introduces the research background,significance and research methods of this research,and puts forward the innovation points of this paper.Researchers of literature review,before the second chapter will focus on the impact of the public opinion on securities prices and stock price prediction based on artificial intelligence research content were summarized,in order to draw lessons from and to improve the research methods.The third chapter introduces the basic theory of the application in this paper,elaborates the method of natural language processing and the model of deep learning model.The fourth chapter shows the entire research process of model building,first of all in the choice of deep learning model recursive neural network(RNN)model types,on the basis of further choose the Long Short-Term Memory model(LSTM),more suitable for research in security prices,the selection of the research object,the data show that the preparation process and the characteristics of data are described,and then illustrates the emotional tendency analysis dictionary compiling method and process,and illustrates the design method,the evaluation index of the final model results and calculation formula are given in detail.Regular stock prices in the fifth chapter shows the research forecast model and the stock price prediction model based on deep learning model two kinds of experimental results,and compares the two models of the experimental results.The sixth chapter summarizes the research of this paper and looks forward the future research direction of this field.This study USES the Chinese a-share market representative 18 stocks in the Shanghai composite index from 2015 to 2017 within three years of experiment data,the data include: stock market data,financial data,public opinion.The public opinion data is different from the general research,using the investment research report of the securities firm and the relevant financial news text data of the media portal.Join up for public opinion research through the influence factors of stock price prediction model is the ability to predict and to join the public opinion influencing factors of stock price prediction model,comparing the predictive power of public opinion are analyzed factors of stock price forecasting model prediction effect is significant.To reflect the role of public opinion information in short term and long term,this study chose the Long Short-Term Memory model(LSTM),which is a branch of recursive neural network(RNN)as the depth of the learning model used in the study.In the study of the index and mean square error(MSE)predict the direction of the accuracy of the deep learning(DS)index to evaluate training model of performance,and to contrast to conventional stock price prediction model and stock price prediction model based on deep learning model results.The final result of this study fully demonstrates that the addition of public opinion information has a certain effect on the prediction effect of securities price prediction model.Study the stock price prediction model based on deep learning model of full consideration to the factors influencing the stock price of public opinion,by using the model for the forecast of stock prices in the actual application will play a role.