Table 5.
Performance results (validation R2≥0.5) of trialled preprocessing and multivariate methods for the prediction of nitrogen in wheat.
Calibration | ||||||
---|---|---|---|---|---|---|
PLSR | PCR | MLR | RF | SVM | ||
SMO | R2 | 0.56 | 0.50 | 0.53 | 0.42 | 0.60 |
RMSE | 0.42 | 0.45 | 0.43 | 0.48 | 0.40 | |
RPD | 1.49 | 1.40 | 1.46 | 1.31 | 1.58 | |
Raw | R2 | 0.59 | 0.49 | 0.48 | 0.42 | 0.62 |
RMSE | 0.41 | 0.45 | 0.45 | 0.47 | 0.39 | |
RPD | 1.53 | 1.38 | 1.39 | 1.32 | 1.60 | |
Validation | ||||||
PLSR | PCR | MLR | RF | SVM | ||
SMO | R2 | 0.59 | 0.54 | 0.57 | 0.33 | 0.43 |
RMSE | 0.41 | 0.44 | 0.42 | 0.52 | 0.48 | |
RPD | 1.56 | 1.47 | 1.53 | 1.22 | 1.33 | |
Raw | R2 | 0.57 | 0.52 | 0.58 | 0.33 | 0.43 |
RMSE | 0.42 | 0.44 | 0.42 | 0.52 | 0.48 | |
RPD | 1.53 | 1.44 | 1.54 | 1.23 | 1.33 |
R2 = coefficient of determination, RMSE; root mean square error; RPD; ratio of performance to deviation. PLSR; partial least-squares regression; PCR; principal components regression; MLR; multiple linear regression; RF; random forest; SVM; support vector machine; SMO; smoothed.