Research On Stock Price Prediction Method Based On Sentiment Analysis And Deep Learning

The volatility of the stock market will have an impact on all aspects of social and economic life.Therefore,effectively predicting the trend of changes in stock prices will not only affect the returns of investors,but also have greater economic and social value.However,the changes in the stock market are affected by many factors,and it is difficult to accurately predict them.Early stock research theories believed that transaction data was not helpful for future predictions.Later,many researchers automatically extracted the laws that predict changes in the stock market based on historical data,but these models cannot capture the laws of non-stationary data.With the development of behavioral finance theory,traditional time series models cannot fully consider the impact of investor psychology and sentiment on the stock market.This paper proposes a stock price prediction method that integrates sentiment analysis and deep learning.The main work includes the following two aspects.Aiming at the problems of stock price prediction methods based on single source data,this paper proposes a stock price prediction method based on multi-source data and sentiment analysis.This method first obtains multiple types of data sources from the Internet and preprocesses them respectively.These data include traditional data sources and non-traditional data sources.Then,the sentiment analysis method based on Convolutional Neural Network(CNN)is used to mine the sentiment information of multi-source data,and the sentiment index of investors is calculated,so as to obtain the investment tendency of investors.Finally,the sentiment index,technical indicators,and historical stock transaction data are used as the feature set of stock prediction,and the long short-term memory(LSTM)network is used to predict the closing price of five listed companies in the Shanghai A-share market in China.The experimental results show that compared with the stock price predicted by a single data source,the stock price predicted by combining multiple data sources is closer to the closing price,and the error result is smaller.Aiming at the unbalanced distribution of text labels in the sentiment analysis method based on CNN neural network,this paper proposes a stock price prediction method based on sentiment analysis and generative adversarial network with the help of game theory.First,a sentiment dictionary database is established in the financial field.Then,the dictionary-based sentiment analysis method is used to calculate the sentiment polarity of financial text data,and calculating the overall sentiment trend of investors every day,that is,the sentiment index.Finally,the generative confrontation network is used to predict the stock market volatility,where the generator generates stock sequence data,and the discriminator uses a convolutional neural network to distinguish the generated data from the real data.This method can dynamically update the prediction results of stocks and obtain smaller error values.