Skip to main content

Table 3 Performance comparison of single benchmark models in all intervals

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

Model

RMSE

MAPE

MAE

DA

One-step ahead

 EMD-LSTM

0.0177

0.0269

0.0123

0.5234

 BiLSTM

0.0159

0.0242

0.0110

0.5390

 ARIMA

0.0153

0.0092

0.0206

0.5431

 LR

0.0943

0.0335

0.0143

0.5277

 SVR

0.0506

0.2179

0.0406

0.5627

 LMH-BiLSTM

0.0071

0.0113

0.0095

0.8167

Two-step ahead

 EMD-LSTM

0.0341

0.0596

0.0266

0.4047

 BiLSTM

0.0290

0.0532

0.0230

0.4676

 ARIMA

0.0242

0.0169

0.2580

0.5132

 LR

0.0327

0.0554

0.0250

0.5037

 SVR

0.0574

0.2196

0.0472

0.5794

 LMH-BiLSTM

0.0073

0.0121

0.0055

0.7855

Three-step ahead

 EMD-LSTM

0.0679

0.1286

0.0573

0.4086

 BiLSTM

0.0395

0.0708

0.0318

0.4554

 ARIMA

0.0245

0.0171

0.2590

0.4836

 LR

0.0420

0.0726

0.0322

0.5473

 SVR

0.0637

0.2214

0.0514

0.5914

 LMH-BiLSTM

0.0080

0.0126

0.0056

0.7654

  1. Bold means the smallest forecasting error among all models