Stock Data Mining Methods And Their Applications Based On Deep Learning

Stock market activities affect the development of the world economy,but also have become a part of people’s daily life.With the digitization of the stock system,the stock market always produces massive stock data.However,because most investors can not analyze the massive and diverse stock data,they are vulnerable to the impact of public opinion in the stock market and lack of professional investment strategies,resulting in heavy losses.Therefore,effective stock data mining can guide the activities of investors and promote the healthy development of the stock marketThis thesis focuses on the methods and applications of stock data mining.For different types of stock data,we study the public opinion analysis method and stock trading strategy model respectively to improve the effectiveness and reliability of the method,and build the "Shanghai Stock Exchange 50" stock data mining system.The main work and contributions of this thesis are as followings:1.To propose a sentiment analysis method based on FinBERT-CNN:aiming at the Guba stock forum(guba.com.cn),this thesis uses FinBERT to learn semantic features,which solves the problem of lack of stock comments annotation data set.At the same time,convolutional neural network(CNN)is used to learn local features.It solves the problem of single feature extraction,and improves the accuracy and effectiveness of the model2.To propose a stock trading strategy model based on deep Q network:aiming at the stock market data and comments,this thesis puts forward the judgment method of stock trading point based on dynamic threshold and the framework of deep Q-network to construct the stock trading strategy.At the same time,the sentiment of Guba stock forum is quantified and integrated into the stock state space representation.It solves the problem that the existing trading strategy model ignores the diversity of stock data,improves the stability of the strategy,and reduces the investment risk3.To construct a "Shanghai Stock Exchange 50" stock data mining system proto-type:based on the demand of massive stock data mining in the market,the prototype of"Shanghai Stock Exchange 50" stock data mining system is realized.The RPA frame-work is applied to design the functions of stock information visualization,stock public opinion display,stock trading point judgment,and portfolio construction,it realized the automatic processing flow of the whole systemHow to use different types of stock data for analysis and mining to assist investors in trading and promote the healthy development of the market is the common goal of research in this field.This thesis studies different types of stock data mining methods Compared with existing research methods,the sentiment analysis method proposed in this paper,as well as the stock trading strategy method,performs better.The "Shanghai Stock Exchange 50" stock data mining system constructed can also be used to assist investors in investment decisions.