The Research On Stock Index Forecasting Based On Deep Learning

Along with the continuous development and improvement of the economy in China,stock investment has become a popular choice for national investment and financial management at this stage.The trend of stock prices not only directly affects the stability of the stock market,but also relates to the sound development of the economy and finance.Investors often need to forecast the trend of stocks in order to obtain benefits.In addition,the government and other relevant departments also need timely and effective supervision and guidance of the market.However,as an extremely complicated dynamic system,stocks will be affected by many factors,among which internal factors include the development trend of the company,and external factors such as macroeconomic development.It is precisely because of the volatilit y,nonlinearity and the lower signal to noise ratio of the stock market that the process of predicting stock price movements is very complicated and difficult.The stock index can reflect the changes in the stock market in a timely and comprehensive manner,from which we can see the trend of the stock price.Therefore,stock index forecasting is important to help investors make accurate decisions.It is also a practical topic common to both the financial and computer fields.The Long-Short term memory neural network(LSTM)is generally used for natural language processing,but some studies have been done on the prediction of time series,which proves that the model can bring certain effects.This paper adds double attention mechanisms(DA)to the LSTM to predict the closing price of the CSI 300 stock index.It consists of an encoder with an input attention mechanism and a decoder with a timeattention mechanism.The input attention mechanism can adaptively select the relevant constituent stocks that ultimately affect the rise or fall of the stock index.The time attention mechanism captures time information for coded inputs over a long period of time.Based on these double attention mechanisms,the double attention mechanismLong-Short term memory neural network model(DA-LSTM)can not only adaptively select the most relevant input features,but also appropriately capture the long term time dependence of time series.In addition,in the selection of the characteristic factors,this paper takes the constituent stocks of the Shanghai and Shenzhen 300 stock indices and the closing price of the stock index as the characteristic factors of the forecasting model,and empirically analyzes the closing price of the stock index.In this paper,the sample index,the selection of experimental data,the composition of the network structure,the selection of the number of hidden layer nodes,the learning rate and the activation function are also discussed.The experimental results show that from the simulation and prediction effects,the constructed DA-LSTM neural network can achieve better prediction results than the LSTM neural network,and can accurately predict the future trend of the stock index in the near future.The model will have some guiding significance in real investment.