Research On Stock Price Forecasting Based On Long Short-term Memory Neural Network And Grey Model

The changing trend of financial market always affects China's economic macro-control and the investment returns of investors and investment institutions.Therefore,more and more scholars try to dig out the hidden laws in financial time series data through different methods,hoping to predict the changing trend of financial market as much as possible.In recent years,the system of China's financial sector has been gradually improved,the financial market has been developing rapidly,and the investment of investors and investment institutions has been increasing.In the context of the demand for information services in the financial sector,the development of computer technology has also provided a technical basis for the demand for financial information.In this paper,LSTM model and GM(1,1)grey model are used as basic tools,and a hybrid model of the two models is constructed to make a detailed study on the stock price prediction.The main research contents are as follows:(1)Select representative stocks from the Shanghai and Shenzhen stock markets in China,collect more than 580,000 pieces of historical data to build a data set,and store them through the My SQL database.(2)Based on LSTM model and GM(1,1)model,the stock price prediction is carried out,and the experimental results show that the above two models perform well in the task of stock closing price prediction.However,if the LSTM model is needed to be used on high stock price samples,it is necessary to make sufficient data preprocessing and model parameter adjustment to obtain better forecast results.(3)In order to further improve the prediction accuracy,the hybrid model of LSTM and GM(1,1)is constructed,and the performance of the hybrid model is tested on the same data set.The experimental results show that the prediction effect and fluctuation amplitude of the hybrid model are better than individual model,and compared with other time series prediction models such as Prophet,quadratic exponential smoothing and ARIMA,the hybrid model also has certain advantages.