Quantitative Analysis Of Stock Market Based On Long Short-Term Memory Network

In the Internet era,people can search for various types of messages on the browser,and the traces of browsing will be recorded by "big data".For example,if you search for "stock" on Baidu,the Baidu index on that day will be recorded once.popularity.Investors' psychological expectations are also reflected in the browser search behavior,that is,behavioral finance.Therefore,it is reasonable to use the Baidu index's search volume to indirectly reflect investors' expectations of the stock market.In the era of big data,machine learning,deep learning and other algorithms are very lively.Most of the previous stock market quantitative analysis is based on the ARMIA time series model.Later studies added technical analysis models such as neural networks,and used long short-term memory network LSTM for stock price trend analysis.The research also appeared in recent years.The long-and short-term memory network RNN is a new type of deep learning time series model based on the cyclic neural network.It has a high degree of self-learning ability and simulation ability,and has the characteristics of memory sustainability.It can predict any step in the future and is more suitable for forecasting financial time series.This paper takes the Yanghe shares of the liquor industry in the A-share market and the opening price forecast of Guizhou Maotai as the research.Firstly,it summarizes the research of behavioral economics and investor attention,summarizes the past research on stock market and keyword hot search,and finds that Baidu index can be used as a keyword hot search variable.Then,the neural network model is described,and the RNN cyclic neural network and the long-short-term memory network model LSTM are compared.It is found that the error value of the long-term memory network model LSTM predicts the stock price is smaller than the cyclic neural network model RNN,so the length is selected.The time-memory network model predicts financial time series more appropriate.Therefore,this paper constructs a prediction model of the opening price of the stock market based on the long and short time memory network,and adds the Baidu index of the stock to the influencing factors.The specific research contents are as follows:First,in the selection of variables,the opening price,closing price,highest price,lowest price,trading volume,price increase,amplitude and Baidu index of t-1 are used as input variables,that is,influencing factors,variables to be predicted.It is the opening price of t day.In the selection of individual stocks,the more representative Yanghe shares in the liquor industry and Guizhou Maotai were selected as research objects,and a forecast model of stock opening price was established.Secondly,in order to compare whether the long-and short-term memory network model is superior to the cyclic neural network model in predicting the opening price,we have established a long-short-term memory network model and a cyclic neural network for the opening price prediction of Yanghe(Kweichow Moutai).Model and measure the predicted effect with the predicted error value.At the same time,in order to compare whether the Baidu index of Yanghe(Kweichow Moutai)will affect the opening price forecast,the input variables of the long-and short-term memory network model are divided into the presence or absence of the Baidu index,and the same applies to the cyclic neural network model.So far,Yanghe shares and Guizhou Maotai have four stock price prediction models.Third,the time span from 2015 to 2018 was used as the study,and the data from 2015 to 2017 was used as the training set and the data in 2018 was used as the test.This paper finds that: 1.The LSTM model is better than the RNN model in predicting the opening model.When using the historical data of t-1 day to predict the opening price of t day,the error values RMSE,MSE and MAE of the long and short memory network model are smaller than the error value of the cyclic neural network prediction,which confirms the LSTM model.The effect is better than the RNN model.Baidu index search volume can predict stock price;2.Whether it is long or short memory network or cyclic neural network model,when the input variable contains Baidu index,the error value in the opening price prediction is lower than the prediction model without Baidu index.This reflects the Baidu index as a variable of people's attention to individual stocks,which has certain practical significance in predicting stock prices.Therefore,the stock market quantitative analysis based on long and short time memory networks has the following meanings:The first is the theoretical significance: the long-and short-time memory network model LSTM has a small error in predicting the stock price,and has certain value for the research accuracy of the stock market.Meanwhile,when we use the Baidu index as the input variable of the model,we find that the prediction effect is better.Use the Baidu Index as a valid input variable.Finally,in practical terms,through the example of Yanghe and Guizhou Moutai,we hope to provide some reference for the stock market investment in the liquor industry.