Research Article
Selection of Machine Learning Models for Oil Price Forecasting: Based on the Dual Attributes of Oil
Table 10
Error with different window sizes (spot prices).
| ā | 40 | 60 | 80 | MAE | RMSE | MAE | RMSE | MAE | RMSE |
| Panel A | PCA-RNN | 0.0499 | 0.0610 | 0.0600 | 0.0723 | 0.1582 | 0.1741 | MDS-RNN | 0.0835 | 0.0971 | 0.0649 | 0.0794 | 0.0486 | 0.0619 | LLE-RNN | 0.0314 | 0.0413 | 0.0323 | 0.0419 | 0.0362 | 0.0479 | RNN | 0.0762 | 0.0879 | 0.0845 | 0.0969 | 0.0439 | 0.0558 |
| Panel B | PCA-LSTM | 0.0994 | 0.1217 | 0.1269 | 0.1461 | 0.0621 | 0.0769 | MDS-LSTM | 0.2225 | 0.2579 | 0.0929 | 0.1126 | 0.1660 | 0.2025 | LLE-LSTM | 0.0399 | 0.0523 | 0.0448 | 0.0587 | 0.0524 | 0.0675 | LSTM | 0.0740 | 0.0908 | 0.0956 | 0.1157 | 0.1217 | 0.1457 |
|
|