Classification And Application Of Stock Emotion Based On Multi-Core Convolutional Neural Network

The Chinese stock market has been developing for more than two decades,but the problems of investors' blindness and high speculation still exist.Among them,how the stock market sentiment acts as an irrational factor in the stock market has always been a hot issue in research.With the development of the network,a large number of investors published their opinions on the future trend of the stock market in the form of comments on social platforms.These comments carried the emotions of the investors and some important information.Coupled with the rapid development of text mining in recent years,it is possible to use the stock review to explore investor sentiment.This paper is mainly based on the text mining technology of deep learning.The stock evaluation is used as the basic data to mine the stock index and quantify.Finally,it is used as the input feature of the stock price trend forecast,and further study the relationship between the emotional index and the stock market.This paper first uses the Scrapy to crawl all the comments of 50 stocks of Oriental Fortune Online Stock,and then selects the data of Shanghai Pudong Development Bank as a sample for data preprocessing,including removing the null value,constructing the dictionary of stock commentary terminology to increase the accuracy of word segmentation,and constructing.Stock market stop words remove text data noise and so on.After each comment is divided,the training of the word embedding model is carried out.This paper uses Word2 vec to train all the pre-processed data training models,and finally selects several stock market terms as the central words and selects the five words closest to them.The results are shown in Table 2.8.The trained word embedding model is then applied to the text classification of the machine learning model and the deep learning model.In this paper,machine learning mainly uses SVM.The deep learning model compares single-core convolutional neural network,multi-core convolutional neural network model and long-short-term memory network model.The results obtained from 483 data as test sets are shown in Table 3.3 and Appendix.2.By comparing the models under different parameters,it is found that the multi-core convolutional neural network performs better than the single-coreconvolutional neural network and the long-short-term memory network,and its accuracy is stable at about 80%,while the other two models have the highest accuracy rate of 73.06%.75.17%.Therefore,this paper uses the multi-core convolutional neural network trained model as the emotion classifier.After adding readings to make the previous emotional indicators more reasonable,this paper will use the convolutional neural network's stock sentiment indicators as the characteristic input to predict the stock price trend,combined with common technical indicators,compared with the previous and the model,the stock sentiment indicators for it is predicted that the stock price trend will have a certain improvement effect,indicating that the model can mine a certain investor sentiment information,which has certain guiding significance for the stock market investment.