Table 1.
The results of discriminant models using full spectra of three different sample sets.
| Sample | Models | Parameters * | Calibration Accuracy (%) | Prediction Accuracy (%) |
|---|---|---|---|---|
| Set 1 | PLS-DA | 16 | 100.00 | 90.91 |
| SVM | (256, 0.3229) | 79.42 | 81.82 | |
| KNN | 4 | 66.17 | 59.09 | |
| SIMCA | (10, 10) | 80.88 | 95.95 | |
| Set 2 | PLS-DA | 6 | 96.67 | 96.67 |
| SVM | (256, 1.7411) | 100.00 | 100.00 | |
| KNN | 4 | 91.11 | 100.00 | |
| SIMCA | (10, 10) | 90.00 | 90.00 | |
| Set 3 | PLS-DA | 15 | 100.00 | 100.00 |
| SVM | (48.5029, 1.7411) | 98.89 | 100.00 | |
| KNN | 3 | 93.33 | 93.33 | |
| SIMCA | (9, 9) | 95.56 | 100.00 |
* The parameters indicate the parameters of each model; the parameters for partial least squares-discriminant analysis (PLS-DA), support vector machine (SVM), k-nearest neighbors (KNN), and soft independent modeling of class analogies (SIMCA) are the optimal number of latent variables, (C, g), the number of nearest neighbors, and number of principal components (PCs) of each class, respectively.