Skip to main content
. 2022 Aug 16;22(16):6121. doi: 10.3390/s22166121

Table 1.

Overall comparison of the models.

Model Structure RMSE MAPE MAE R2 Time (s)
LSTM 1 Deeper Layer 2.35×10−2 6.60 ×10−2 1.56 ×10−3 0.7104 75.86
2 Deeper Layers 3.01 ×10−2 8.07 ×10−2 1.95 ×10−2 0.5269 107.22
3 Deeper Layers 4.49 ×10−2 1.37 ×10−1 3.17 ×10−2 0.0506 155.01
4 Deeper Layers 5.22 ×10−2 1.63 ×10−1 3.75 ×10−2 0.4208 182.07
GMDH 1 Max. Layer 4.79 ×10−3 3.32 ×10−3 7.18 ×10−4 0.9880 2.21
2 Max. Layers 4.35 ×10−3 1.90 ×10−3 4.70 ×10−4 0.9901 2.90
3 Max. Layers 5.10 ×10−3 9.06 ×10−3 2.10 ×10−3 0.9864 1.92
4 Max. Layers 6.00 ×10−2 3.11 ×10−2 7.30 ×10−3 0.8819 4.68
ANFIS FCM 1.15 ×10−2 2.38 ×10−2 4.85 ×10−3 0.9304 25.08
Grid Partitioning 5.23 ×10−3 5.68 ×10−3 1.45 ×10−3 0.9857 92.53
Subt. Clustering 4.48 ×10−3 3.29 ×10−5 3.43 ×10−5 0.9895 73.75
Ensemble Bagging 4.19 ×10−3 1.88 ×10−3 3.16 ×10−4 0.9909 4443.30
Boosting 3.40 ×10−2 1.27 ×10−1 2.73 ×10−2 0.3971 31,874.70
Random Subsp. 4.94 ×10−3 1.09 ×10−2 2.21 ×10−3 0.9872 25,155.34
Stacking 5.04 ×10−2 1.56 ×10−1 3.59 ×10−2 0.3255 1295.48