Forecasting gold price using a novel hybrid model with ICEEMDAN and LSTM-CNN-CBAM
Yanhui Liang,Yu Lin,Qin Lu
DOI: https://doi.org/10.1016/j.eswa.2022.117847
IF: 8.5
2022-11-15
Expert Systems with Applications
Abstract:Gold price has always played an important role in the world economy and finance. In order to predict the gold price more accurately, this paper proposes a novel decomposition-ensemble model. Firstly, the original gold prices are decomposed into sublayers with different frequencies by the improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN). Secondly, the long short-term memory, convolutional neural networks, and convolutional block attention module (LSTM-CNN-CBAM) joint forecasting all sublayers. Finally, the prediction of the sublayers with different models is reconstructed as the final predicted results with the summation method. Among them, the proposed model could capture the essence of sequence effectively through ICEEMDAN algorithm, extract the long-term effect of the gold price by LSTM, mining the deep features of gold price data with CNN, and improving the feature extraction ability of the network through CBAM. It is proved by experiment that the cooperation among LSTM, CNN and CBAM can strengthen the modeling ability and improve the prediction accuracy. Moreover, the decomposition algorithm ICEEMDAN can further increase the forecast precision, and the prediction effect is better than other decomposition methods. Overall, the novel model ICEEMDAN-LSTM-CNN-CBAM (ILCC) could enhance the prediction accuracy and outperform other related comparative models.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science