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

Financial Innovation

Table 4 Performance comparison of hybrid benchmark models in all intervals

Model	RMSE	MAPE	MAE	DA
One-step ahead
L-BiLSTM	0.0451	0.0821	0.0362	0.5155
M-BiLSTM	0.0210	0.0345	0.0161	0.6146
H-BiLSTM	0.0121	0.0215	0.0090	0.6618
LM-BiLSTM	0.0360	0.0588	0.0274	0.6547
LH-BiLSTM	0.0163	0.0305	0.0138	0.6426
MH-BiLSTM	0.0143	0.0269	0.0112	0.7609
LMH-BiLSTM	0.0074	0.0185	0.0092	0.8172
Two-step ahead
L-BiLSTM	0.0569	0.0975	0.0416	0.4715
M-BiLSTM	0.0253	0.0439	0.0202	0.6188
H-BiLSTM	0.0122	0.0203	0.0090	0.6859
LM-BiLSTM	0.0338	0.0554	0.0253	0.6506
LH-BiLSTM	0.0146	0.0266	0.0115	0.6341
MH-BiLSTM	0.0095	0.0168	0.0073	0.7497
LMH-BiLSTM	0.0076	0.0118	0.0054	0.7852
Three-step ahead
L-BiLSTM	0.0698	0.1336	0.0568	0.4519
M-BiLSTM	0.0324	0.0561	0.0254	0.5670
H-BiLSTM	0.0139	0.0257	0.0108	0.6270
LM-BiLSTM	0.0343	0.0651	0.0284	0.6466
LH-BiLSTM	0.0188	0.0362	0.0159	0.5984
MH-BiLSTM	0.0090	0.0170	0.0072	0.7175
LMH-BiLSTM	0.0078	0.0121	0.0050	0.7657