Research On Stock Market Sentiment And Trend Analysis Model Based On Deep Learning

With the continuous advancement of Chinese natural language processing technology,combined with the basic judgment that the domestic stock market is in a weakly efficient market,the research on judging market sentiment by analyzing various text data released on the day is gradually increasing,most of which are through the analysis of news text sentiment Tendency to judge market sentiment,and there are not many studies on judging market sentiment by analyzing and commenting on sentiment orientation.In order to analyze the relationship between the sentiment tendency of stock reviews and market trends,first of all,in response to the lack of comment data sets,crawler technology is used to crawl the comments of relevant forums,and perform emotion labeling,text cleaning and other tasks.Then compare the Word2 vec model and the ALBERT model,select the ALBERT model for text vectorization,build the ALBERT-CNN-Bi GRU model,adjust the model parameters and compare with the CNN and Bi GRU models.The model is used to predict the sentiment tendency of stock reviews,to explore the relationship between stock comment sentiment and stock trends,and to analyze the sentiment trend of comments to judge the market position.The research results show that the ALBERT-CNN-Bi GRU model has a better effect on the sentiment classification of comments,reaching an accuracy of 85.99% on the test set,which is higher than the accuracy of 80.92% and 82.34% of the CNN and Bi GRU models.Analyze and calculate the sentiment tendency of the comments to get the sentiment index of the day,through which the position of the bottom of the market can be better judged.The accuracy rate of the Shanghai Stock Exchange Index combined with this indicator and the LSTM model trained by the ups and downs prediction model reached 53.88%,which is 8.73%higher than the accuracy rate without this indicator.