Hybrid data decomposition-based deep learning for Bitcoin prediction and algorithm trading

Financial Innovation

Table 5 Horizontal performance comparisons

Model	RMSE	MAPE	MAE	DA
One-step ahead
LMH-BiLSTM	0.0182	0.0230	0.0115	0.8170
EMD-LSTM	0.0534	0.0611	0.0783	0.6117
LSTM (Altan et al. 2019)	0.0668	0.0732	0.1065	0.5409
LR (Cohen 2020)	0.0646	0.0770	0.1235	0.5155
ARIMA (Wirawan et al. 2019)	0.0663	0.0821	0.1308	0.5719
Two-step ahead
LMH-BiLSTM	0.0075	0.0118	0.0050	0.7852
EMD-LSTM	0.0487	0.0613	0.0637	0.6703
LSTM (Altan et al. 2019)	0.0522	0.0637	0.0914	0.5741
LR (Cohen 2020)	0.0564	0.0616	0.0844	0.5686
ARIMA (Wirawan et al. 2019)	0.0640	0.0758	0.1073	0.6289
Three-step ahead
LMH-BiLSTM	0.0072	0.0126	0.0056	0.7652
EMD-LSTM	0.0475	0.0620	0.0576	0.6723
LSTM (Altan et al. 2019)	0.0668	0.0792	0.0688	0.6583
LR (Cohen 2020)	0.0587	0.0695	0.0670	0.6527
ARIMA (Wirawan et al. 2019)	0.0537	0.0823	0.0767	0.6203