Application Of LSTM Hybrid Model In Shanghai Composite Index Forecast

With the continuous improvement of China’s reform and opening up level,the stock market,as an important direct financing channel in China,has become more and more important in the economic operation.Try to establish a more accurate and comprehensive stock price forecasting model can provide a reference for the investment operations of the majority of stock investors in China,so that the stock market can better serve the real economy.On the other hand,it can investigate the risks and fluctuations of China’s financial market and economic operation,which is of great significance in the formulation of relevant financial policies and macro-control.This article selects the Shanghai Composite Index to represent the overall trend of the Chinese stock market and selects 15 relevant indicator variables that affect stock prices in three aspects:historical transaction information,macroeconomics and investor sentiment.After preprocessing each indicator variable,we input it into the LSTM model and establish two hybrid models:PCA-LSTM model and XGB-LSTM model.Then we train the data from January 2,2014 to July 31,2020 and predict the closing index of the Shanghai Composite Index from August 3,2020(Monday)to December 31,2020.Finally,we use the grid search method to optimize the hyperparameters of the two hybrid models and compare the prediction effects of the two models.The results show that the adjusted PCA-LSTM hybrid model has less error in forecasting data and better forecasting ability.It can predict the closing index of the Shanghai Composite Index more accurately and comprehensively.It also provides new research ideas for the follow-up research on the prediction of high-dimensional time series.