A Hybrid Attention-Based EMD-LSTM Model for Financial Time Series Prediction

In order to improve the accuracy of financial time series prediction, a hybrid model is proposed in this paper which consists of the empirical mode decomposition (EMD) and the attention-based long short-term memory (LSTM-ATTE). EMD can effectively decompose financial time series into many inherent mode functions (IMFs) of multiple levels and input these IMFs into LSTM-ATTE for prediction. The attention mechanism can adaptively extract input features of the IMF and improve the accuracy of the LSTM-ATTE prediction. Finally, the predicted results are combined to obtain the final predicted results. The predictive performance of the proposed model is verified by linear regression analysis of the stock market index. In addition, by comparing the prediction results with other models, the proposed model has better performance in prediction accuracy.