Table 1.
Predictive capabilities of testing dataset from the Arid and Semi-arid Agriculture Institute of China (ASAIC) preprocessing by different method and modeling using the Partial least squares (PLS) or BP neural network (BPNN): correlation coefficient (R2), and root mean square error (RMSE).
| Preprocessing Method | Model | R2 | RMSE |
|---|---|---|---|
| MSC | BPNN | 0.8256 | 1.9785 |
| SNV | BPNN | 0.8141 | 1.8945 |
| MSC | PLS | 0.8127 | 1.7269 |
| SNV | PLS | 0.8116 | 1.7996 |
| SWS-MSC | BPNN | 0.8482 | 1.7940 |
| SWS-SNV | BPNN | 0.8454 | 1.7970 |
| SWS-MSC | PLS | 0.8429 | 1.7369 |
| SWS-SNV | PLS | 0.8492 | 1.7216 |