Research Article
Short-Term Daily Univariate Streamflow Forecasting Using Deep Learning Models
Table 4
Performance comparison of the proposed models for different input time lags in Borkena station.
| Borkena | T + 1 | T + 2 | T + 3 | T + 4 | RMSE | MAE | MAPE | R2 | TTPE (sec) | RMSE | MAE | MAPE | R2 | TTPE (sec) | RMSE | MAE | MAPE | R2 | TTPE (sec) | RMSE | MAE | MAPE | R2 | TTPE (sec) |
| MLP | 1.09 | 0.47 | 0.85 | 0.98 | 0.24 | 0.89 | 0.38 | 0.76 | 0.99 | 0.16 | 1.06 | 0.35 | 0.19 | 0.98 | 0.12 | 0.81 | 0.29 | 0.34 | 0.99 | 0.29 | GRU | 1.15 | 0.64 | 0.84 | 0.98 | 1.29 | 0.85 | 0.31 | 0.16 | 0.99 | 0.79 | 0.91 | 0.35 | 0.18 | 0.98 | 0.99 | 1.35 | 0.69 | 1.41 | 0.97 | 1.31 | S-LSTM | 1.07 | 0.38 | 0.29 | 0.98 | 0.30 | 0.87 | 0.36 | 0.26 | 0.99 | 0.54 | 1.02 | 0.64 | 0.99 | 0.98 | 0.81 | 1.53 | 0.96 | 1.72 | 0.96 | 0.58 | Bi-LSTM | 1.06 | 0.39 | 0.31 | 0.98 | 0.88 | 0.90 | 0.46 | 0.76 | 0.99 | 1.55 | 1.12 | 0.51 | 0.95 | 0.98 | 1.55 | 0.98 | 0.44 | 0.53 | 0.98 | 5.3 |
|
|
TTPE(training time per epochs). The bold values indicate the best performance score for each time lag. |