Table 3.
Best results of predicting ADA levels in WF using MEWs based on the different numbers of WF samples in training set and validation set.
| NIR Source (NT/V Group) |
Spectra | Number of MEWs |
Modeling Algorithm |
Quantitative Regression Analysis | Discriminant Analysis (Whether Excessive ADA) | |||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| LVs | Training Set | Validation Set | LVs | Training Set | Validation Set | |||||||||||
| R2T | RMSET | R2V | RMSEV | RPD | R2T | RMSET | AOC | R2V | RMSEV | AOC | ||||||
| NIR (NT/V = 168/42) |
SG1D | 24 | PLS | 4 | 0.9810 | 4.18 | 0.9819 | 4.08 | 7.4889 | 3 | 0.9091 | 0.30 | 98.21% | 0.9119 | 0.30 | 97.62% |
| MLR | - | 0.9914 | 2.81 | 0.9898 | 3.06 | 9.9001 | - | 0.9410 | 0.24 | 99.40% | 0.9369 | 0.25 | 100% | |||
| HSI-NIR (NT/V = 147/63) |
RAW | 23 | PLS | 7 | 0.9892 | 3.15 | 0.9835 | 3.88 | 7.7972 | 8 | 0.9038 | 0.31 | 98.64% | 0.8765 | 0.35 | 100% |
| MLR | - | 0.9919 | 2.72 | 0.9837 | 3.86 | 7.8388 | - | 0.9101 | 0.30 | 99.32% | 0.8787 | 0.35 | 100% | |||