Table 5 Statistical results for each degree of polynomial kernel function of SVR in estimation of cloudless DLR and the statistically significant models are in italics
From: Deep learning and regression modelling of cloudless downward longwave radiation
Degree | Training phase (n = 143) | Testing phase (n = 62) | ||||||
|---|---|---|---|---|---|---|---|---|
r | RMSE | MBE | ts | r | RMSE | MBE | ts | |
1 | 0.8238 | 11.7369 | 1.3666 | 1.0951 | 0.7738 | 12.1464 | 1.3254 | 0.9880 |
2 | 0.8307 | 11.6079 | 1.7796 | 1.3963 | 0.7872 | 11.8693 | 1.9158 | 1.4720 |
3 | 0.8328 | 11.5971 | 2.0420 | 1.6099 | 0.7727 | 12.3154 | 1.9494 | 1.4428 |
4 | 0.8322 | 11.5120 | 1.3902 | 1.0949 | 0.7711 | 12.1809 | 1.6670 | 1.2434 |
5 | 0.8407 | 11.2633 | 1.5587 | 1.2576 | 0.7621 | 12.5872 | 2.0242 | 1.4664 |
6 | 0.8515 | 10.8724 | 1.3523 | 1.1282 | 0.7728 | 12.1841 | 0.9570 | 0.7091 |
7 | 0.7682 | 13.3322 | − 2.0421 | 1.3950 | 0.5689 | 23.4873 | − 4.3354 | 1.6903 |
8 | 0.7929 | 12.5475 | − 0.0403 | 0.0281 | 0.3595 | 57.8330 | − 8.9456 | 1.4091 |
9 | − 0.2772 | 102.237 | − 27.5245 | 2.5159 | 0.1523 | 122.967 | − 62.3317 | 5.2624 |
10 | 0.0539 | 1333.76 | − 1230.32 | 21.5006 | 0.0319 | 1256.93 | − 1216.31 | 34.5369 |
11 | 0.2260 | 182.977 | − 177.028 | 34.4277 | 0.3211 | 185.898 | − 181.001 | 38.4313 |