TABLE 2.
Accuracy of KNN-cosine algorithm as a function of cutoff wavelength. a
| Wavelength (nm) | Error (%) | |||||
|---|---|---|---|---|---|---|
| Cutoff (nm) | WL1 | WL2 | WL3 | Ordered | Disordered | Global |
| 175 | 214 | 218 | 232 | 1.6 | 0 | 1.3 |
| 176 | 214 | 218 | 232 | 1.5 | 0 | 1.3 |
| 177 | 214 | 218 | 232 | 1.5 | 0 | 1.3 |
| 178 | 214 | 218 | 232 | 1.5 | 0 | 1.3 |
| 179 | 214 | 218 | 232 | 1.5 | 0 | 1.3 |
| 180 | 197 | 206 | 233 | 4 | 0 | 3.4 |
| 183 | 197 | 206 | 233 | 4 | 0 | 3.3 |
| 185 | 197 | 206 | 233 | 3.9 | 0 | 3.1 |
| 190 | 197 | 206 | 233 | 4.7 | 1.7 | 3.9 |
| 195 | 197 | 206 | 233 | 4.7 | 1.5 | 3.8 |
| 198 | 198 | 205 | 237 | 4 | 2.9 | 3.7 |
| 200 | 212 | 217 | 225 | 3.3 | 7.5 | 4.6 |
| 205 | 212 | 217 | 225 | 3.3 | 7.4 | 4.6 |
a
Wavelengths of the data points (WL1, WL2, WL3) for the best performance at each cutoff and the errors of classification are shown.