Table 1 Comparison between proposed method and previous research.
Author | Location | Methods | Period | R² | MSE | RMSE | MAE |
|---|---|---|---|---|---|---|---|
Sareen et al.11 | Gujarat, India | KNN + CEEMDAN + BiLSTM | 1-hours | 94% | NA | 0.41 | 0.31 |
Bommidi et al.2 | Block Island and Texas | Transformer + ICEEMDAN | 48-hours | 90% | NA | 0.75 | NA |
Hu et al.6 | western mountains of Chongqing, China | Autoencoder + VMD + Optimized fuzzy mapping network | 144-hours | 98% | NA | NA | 0.462 |
Liang et al.12 | Guangzhou | LSTM + VMD | 5-hours | NA | NA | 0.112 | 0.086 |
Wang et al.10 | Denver | Transformer + RF feature selection | 36-hours | 44% | 0.52 | ||
Zhang et al.13 | China | CNN + BiLSTM | 1-day | NA | NA | 0.77 | 0.56 |
Lin et al.14 | Fujian | GAOformer | 2-hours | NA | 3.85 | NA | 1.45 |
Li et al.15 | Arctic region | CNN-LSTM + CEEMDAN | 16-hours | NA | 0.3960 | 0.6293 | NA |
Bashir et al.3 | Pakistan | Seq-2-Seq + Harris hawk’s | 2-days | NA | NA | 0.639 | 0.474 |
Chen et al.4 | China | CNN-LSTM-Autoencoder | 30-minutes | NA | NA | 0.34 | 0.25 |
Jiang et al.16 | Shenzhen | Graph Neural Network + Temporal Convolutional Network + VMD | 12-hours | 85% | NA | 0.39 | NA |
Chen et al.17 | Karamay | Hilbert–Huang + Nonlinear Autoregressive Dynamic Neural Network | 1-day | 90% | NA | 1.99 | NA |
Houndekindo et al.18 | Canada | RF + Gradient Boosting | 1-hour | NA | NA | 1.47 | 1.13 |
Yu et al.19 | Southern Mississippi | CNN + RNN | 30-minutes | NA | NA | 1.71 | 1.32 |
Jiang et al.20 | Shandong Province, China | CGRU + XGBoost | 2-hours | NA | NA | 0.74 | 0.53 |
Yuan et al.21 | NA | AdaBoost + Relevance Vector Machine | 15-minutes | 95% | NA | 10.403 | NA |
Zeng et al.22 | Average of Six Locations | LightGBM + ANN | 1-hour | 97% | NA | 23.02 | 10.55 |
Proposed | Zaafarana, Egypt | LightGBM + VMD | 1-month | 98% | 0.02 | 0.15 | 0.12 |
