Design And Implementation Of Stock Investment Model Based On Text Analysis And Reinforcement Learning Technology

By the end of 2020,there are 4140 listed companies in China’s A-share market,with a total market value of more than 86 trillion yuan.With the wave of nationwide speculation at the end of the year,a large number of investors are attracted by the capital market,and the popularity of social media also makes the financial public opinion data on the Internet soar.How to use data mining and other artificial intelligence tools to extract and analyze effective information from the network media platform and derive effective factors and strategies has become a hot research topic of securities companies in recent years.This paper systematically compares the academic research on text mining and deep reinforcement learning at home and abroad,introduces the theoretical basis of related algorithms in detail,and fully demonstrates the feasibility of applying artificial intelligence technology and deep learning algorithm to the field of investment,supplemented by reinforcement learning algorithm.At the same time,we make up for and improve the deficiency of genetic programming optimization and low efficiency of traditional reinforcement learning in domestic literature.First of all,this paper uses crawler technology to crawl stock related microblog,obtains the stock comment text data,uses snow NLPtechnology to quantify the stock comment information into emotional factors,and combines with statistics to construct six basic factors.Then,through genetic programming,with mutual information and excess return,20 public opinion factors with higher return are mined from the six most basic factors.Then,the LSTM based on attention mechanism uses 20 public opinion factors to predict the rise and fall of stocks,and measures the return rate of trading by buying stocks with higher probability of rise and selling stocks with lower probability of rise.By comparing the return rate,it evaluates the stock selection methods of different machine learning algorithms.Finally,the LSTM based on attention mechanism has the highest return rate,The best prediction result is obtained Design and implementation of stock investment model based on text analysis and reinforcement learning technologySecondly,considering Eliot’s wave theory and Markov state,an optimized deep reinforcement learning method is proposed to overcome the shortcomings of the original algorithm.The traditional wave theory method is improved.Firstly,the K-line data is processed by inclusion relation,and then it is divided into top classification and bottom classification.Then the high-order classification finds out the high and low points of each band.Taking the interval between the secondary high and the secondary low points as the center,the stock price is designed into three intervals: below the center,inside the center and above the center,and the three intervals are represented by numbers 0,1 and 2,The three intervals are set to the state of Markov chain in reinforcement learning,and the probability transfer matrix of each state is calculated.LSTM based on attention mechanism uses 20 public opinion factors to predict the rise and fall of stocks.When the predicted probability of stock rise is large and the predicted Markov state is 0,it buys stocks,When the predicted Markov state is 1,the position will be locked and the stock will not be traded.The predicted stock will have a higher probability of decline,and when the predicted Markov state is 2,the position will be cleared and the stock will be sold.With the help of reinforcement learning strategy gradient method,this method takes the useful features as the basis,considers the parameters in the neural network,analyzes the prediction ability of future learning,and perfectly integrates with the decision-making ability of reinforcement learning.When facing a large number of historical data,the financial quantitative trading model and trading method based on deep reinforcement learning can ensure that the intelligent trading algorithm can carry out data mining in the real-time stock trading.At the same time,investors can accurately grasp the trading point,complete the correct investment judgment,and reduce the risk in the whole investment process.Finally,in order to solve the shortcomings of traditional quantitative trading software,a quantitative trading system with deep reinforcement learning is designed.The system realizes the purpose function of intelligent decision-making based on stock,and has strong application value.A series of experiments show that the model designed in this paper has achieved good results in both public opinion analysis and investment decision analysis,and also verifies the effectiveness of the proposed model and the correctness of this paper.Based on the deep reinforcement learning technology,and combined with the usual financial quantitative transaction analysis method,it is conducive to give full play to the advantages of both,further promote the cross integration of artificial intelligence,deep learning algorithm,economics and management,and provide reference for responding to and promoting the practice of the national artificial intelligence strategy.