Research On Financial Time Series Prediction Based On EMD-SAE-LSTM Model

The prediction of financial data has always been a challenging problem in various fields such as finance,statistics and mathematics.In recent years,with the development of information technology,the massive retention of financial data and the in-depth study of AI algorithms,more and more scholars began to develop artificial intelligence algorithms in the field of financial time series prediction.Deep learning technology,as a hot topic of artificial intelligence,has been gradually applied in the financial field by scholars.This paper proposes a new deep learning stock price prediction framework: EMD-SAE-LSTM model.EMD is used to decompose the original financial time series.SAE is used to extract the deep features of decomposed data.LSTM is used to predict stock prices.In this paper,the daily closing price of stock index and the minute price of individual stock are forecasted.In this paper,the closing price prediction experiment of stock price index is carried out on the daily data sets of Shanghai Composite Index,Shenzhen Composite Index,SZSE Component Index and Shanghai & Shenzhen 300 Index.In order to determine the feature dimension selected by SAE,four structures were selected to conduct experiments using the LSTM model MAPE as evaluation criteria,and the feature dimension which could make the best prediction effect of LSTM was selected.Then,the model presented in this paper is compared with the cyclic neural network(RNN),LSTM and EMD-LSTM by R-Square and MAPE.The results show that the model has a strong ability to predict the daily closing price of stocks,which verifies the effectiveness of the model presented in this paper.In this paper,15-minute,30-minute and 60-minute forward forecasting models are established on the minute data set of China Ping An Stock Price Index to investigate the forecasting ability of this model on individual stock prices.The results show that the model performs best when 60 minutes of data are used to predict the next minute's stock price.At the same time,compared with RNN,LSTM,EMD-LSTM models by MAE,MAPE,RMSE and R-Square,the results show that the model proposed in this paper has the best performance in the above four indicators,which verifies the prediction ability of this model on the minute price of individual stocks.