Table 8.
Modelling performance of PLSDA and SVM classifiers built from one concentration and applied to other concentrations (deposited on Al) using 3500–2600 cm−1.
| Applied to | 10 OD | 1 OD | 0.1 OD | ||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Built from | LVs | OA | MCC | Sen | Spe | OA | MCC | Sen | Spe | OA | MCC | Sen | Spe | ||
| PLSDA | 10 OD | 4 | 99% | 0.98 | 0.99 | 0.98 | 87% | 0.72 | 0.83 | 0.89 | 55% | 0.16 | 0.78 | 0.38 | |
| 1 OD | 9 | 96% | 0.91 | 0.97 | 0.95 | 100% | 1.00 | 1.00 | 1.00 | 74% | 0.46 | 0.67 | 0.79 | ||
| 0.1 OD | 5 | 57% | 0.14 | 0.11 | 0.96 | 72% | 0.42 | 0.56 | 0.84 | 100% | 1.00 | 1.00 | 1.00 | ||
| SVM | 10 OD | - | 100% | 1.00 | 1.00 | 1.00 | 92% | 0.85 | 1.00 | 0.86 | 62% | 0.28 | 0.78 | 0.50 | |
| 1 OD | - | 96% | 0.92 | 0.95 | 0.96 | 100% | 1.00 | 1.00 | 1.00 | 50% | 0.08 | 0.78 | 0.29 | ||
| 0.1 OD | - | 57% | 0.17 | 0.09 | 0.99 | 55% | 0.06 | 0.44 | 0.63 | 100% | 1.00 | 1.00 | 1.00 | ||
OA: overall accuracy; MCC: Matthews correlation coefficient; Sen: sensitivity; Spe: specificity.