Table 3.
Discrimination results of different models for differentiating WT and CRISPR/Cas9-induced rice mutants based on full wavelengths and optimal wavelengths.
| Rice varieties | Discrimination model | Model build on full wavelength | Model build on optimal wavelength selected by SPA | Model build on optimal wavelength selected by PCA-loadings | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Parametera | Calibration set | Prediction set | Parameter | Calibration set | Prediction set | Parameter | Calibration set | Prediction set | ||
| Huaidao-1 | SVM | 256, 3.0314 | 92.38% | 92.5% | 256, 48.5029 | 93% | 92.75% | 256, 84.4485 | 86.13% | 81% |
| ELM | 28 | 90.25% | 93.25% | 9 | 91.37% | 92% | 64 | 83.38% | 81.25% | |
| Nanjing46 | SVM | 256, 3.0314 | 89.75% | 88% | 256, 84.4485 | 91.25% | 89.50% | 256, 27.8576 | 80.5%% | 76.50% |
| ELM | 35 | 88.62% | 90.25% | 14 | 88.75% | 90% | 35 | 80.13% | 80.75% | |
apar indicates the parameters of the discrimination models, (c,g) for SVM, the optimum number of hidden nodes for ELM.