Research On Stock Price Trend Prediction Based On LSTM Neural Network

As an important financial instrument,stocks have become more and more mature.The scale of stock trading markets in various countries has grown rapidly,and the transaction amount and transaction frequency have kept on record.Stocks are an effective means for listed companies to raise funds and make profits from public investment,while stock prices are considered a barometer of listed companies.In view of this,the academic and industry research on stock price forecasting and risk control has been quite enthusiastic for a long time.It is noted that the daily fluctuations of stock prices are presented by a combination of factors.It is generally not recommended to use classical statistical models and methods such as linear regression and stationary time series analysis directly for trend forecasting of stock prices.Aiming at the characteristics of stock price data,overcoming the shortcomings of traditional forecasting models and exploring new effective forecasting models can help boost investors'market confidence and create a more active and healthy capital market,which has certain theoretical significance and application value.The main work of the thesis is:1.The principle of pre-back propagation of RNN neural network and long-term and short-term memory neural network(LSTM)is introduced.The defects of using RNN neural network to predict stock price and the use of long-and short-term memory neural network(LSTM)to establish stock forecast are theoretically expounded.The rationality of the model.2.Using the stock data of CSI 300(000300)to train the long-short-term memory network(LSTM)model built in this paper,including the selection of the model input layer,the number of output layer nodes,the selection of the number of hidden layers and the number of nodes,and the optimization algorithm.,learning rate,over-fitting method,and choice of input days.3.In order to verify that the long-and short-term memory neural network(LSTM)model built in this paper has excellent performance in predicting stock price trends.The opening price,closing price and closing price of the two stocks of the Shanghai Composite Index(000001)and Shenzhen Composite Index(399106)from February 1,2010 to January 31,2019 were selected.,the lowest price(Low)and volume(Volume)data comparison analysis of the five long-term memory neural network(LSTM)model and RNN neural network model for the prediction performance of the opening price,through the stock price chart and three The performance evaluation indicators RRMSE(relative root mean square error),MAE(mean absolute error),and MAPE(average absolute error percentage)confirm that the long-and short-term memory neural network(LSTM)model built in this paper is excellent in predicting stock price trends.Performance.