Table 5 Performance of soil free iron content models based on spectrum transforms, variable selection, modeling methods *.
From: Estimation of soil free Iron content using spectral reflectance and machine learning algorithms
Spectrum transform, variable selection, and modeling method | Training | Testing | ||||
|---|---|---|---|---|---|---|
R2 | RMSE (g/kg) | RRMSE (%) | R2 | RMSE (g/kg) | RRMSE (%) | |
Original | 0.725 | 6.145 | 27.9 | 0.567 | 7.578 | 34.7 |
FD | 0.797 | 5.550 | 25.1 | 0.592 | 7.581 | 34.7 |
SNV | 0.710 | 6.305 | 28.9 | 0.579 | 7.523 | 34.4 |
CR | 0.793 | 5.371 | 24.4 | 0.644 | 7.14 | 32.7 |
Full | 0.750 | 5.890 | 26.8 | 0.577 | 7.607 | 34.9 |
Correlated | 0.698 | 6.379 | 29.0 | 0.517 | 8.015 | 36.6 |
PCA | 0.821 | 5.260 | 23.9 | 0.692 | 6.744 | 30.9 |
PLS | 0.640 | 7.089 | 31.8 | 0.588 | 7.261 | 33.6 |
SVM | 0.701 | 6.427 | 30.4 | 0.603 | 7.414 | 33.9 |
RF | 0.926 | 3.863 | 17.5 | 0.520 | 8.282 | 37.7 |
DNN | 0.720 | 6.330 | 28.3 | 0.646 | 6.976 | 31.9 |
