Table 2. Species classification results for the classification of the validation set using different analysis methods and the one-vs-all approach.
| Class 1 | Class 2 | Number of variables | Settings | Individuals correct prior to exclusion | Individuals incorrect prior to exclusion | Exclusion range | Individuals correct after exclusion | Individuals incorrect after exclusion | Nests correct after exclusion | |
|---|---|---|---|---|---|---|---|---|---|---|
| PLS | T. alpestre | T. caespitum/impurum/sp. B | 1,801 | 12 | 26 (57.8%) | 19 (42.2%) | 1.25–1.75 | 21 (46.7%) | 0 (0.0%) | 10 (66.7%) |
| T. caespitum | T. alpestre/impurum/sp. B | 1,801 | 10 | 20 (43.4%) | 25 (56.6%) | 1.20–1.80 | 3 (6.7%) | 0 (0.0%) | 3 (20.0%) | |
| T. impurum | T. alpestre/caespitum/sp. B | 1,801 | 15 | 20 (43.4%) | 25 (56.6%) | 1.20–1.80 | 8 (17.8%) | 0 (0.0%) | 7 (46.7%) | |
| T. sp. B | T. alpestre/caespitum/impurum | 1,801 | 10 | 9 (20.0%) | 36 (80.0%) | 1.35–1.65 | 2 (4.4%) | 0 (0.0%) | 2 (13.3%) | |
| Total | 75 (41.7%) | 105 (58.3%) | 34 (18.9%) | 0 (0.0%) | 22 (36.7%) | |||||
| ANN | T. alpestre | T. caespitum/impurum/sp. B | 150 | 5 | 35 (77.8%) | 10 (22.2%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) |
| T. caespitum | T. alpestre/impurum/sp. B | 150 | 23 | 19 (42.2%) | 26 (57.8%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| T. impurum | T. alpestre/caespitum/sp. B | 150 | 34 | 25 (55.6%) | 20 (44.4%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| T. sp. B | T. alpestre/caespitum/impurum | 150 | 18 | 24 (53.3%) | 21 (46.7%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| Total | 103 (57.2%) | 77 (42.8%) | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |||||
| RF | T. alpestre | T. caespitum/impurum/sp. B | 150 | 12 | 3 (6.7%) | 42 (93.3%) | 0.35–0.65 | 2 (4.4%) | 0 (0.0%) | 2 (13.3%) |
| T. caespitum | T. alpestre/impurum/sp. B | 150 | 12 | 4 (8.9%) | 41 (91.1%) | 0.10–0.90 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| T. impurum | T. alpestre/caespitum/sp. B | 150 | 12 | 3 (6.7%) | 42 (93.3%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| T. sp. B | T. alpestre/caespitum/impurum | 150 | 12 | 4 (8.9%) | 41 (91.1%) | 0.20–0.80 | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| Total | 14 (7.8%) | 166 (92.2%) | 2 (1.1%) | 0 (0.0%) | 2 (3.3%) | |||||
| T. alpestre | T. caespitum/impurum/sp. B | 1,801 | 42 | 2 (4.4%) | 43 (95.6%) | 0.30–0.70 | 2 (4.4%) | 0 (0.0%) | 2 (13.3%) | |
| T. caespitum | T. alpestre/impurum/sp. B | 1,801 | 42 | 13 (28.9%) | 32 (71.1%) | 0.15–0.85 | 2 (4.4%) | 0 (0.0%) | 2 (13.3%) | |
| T. impurum | T. alpestre/caespitum/sp. B | 1,801 | 42 | 14 (31.1%) | 31 (68.9%) | – | 0 (0.0%) | 0 (0.0%) | 0 (0.0%) | |
| T. sp. B | T. alpestre/caespitum/impurum | 1,801 | 42 | 9 (20.0%) | 36 (80.0%) | 0.25–0.75 | 2 (4.4%) | 0 (0.0%) | 2 (13.3%) | |
| Total | 38 (21.1%) | 142 (78.9%) | 6 (3.3%) | 0 (0.0%) | 6 (10.0%) |
Notes.
- PLS
- Partial least squares regression
- ANN
- Artificial neural networks
- RF
- Random forests
- Settings
- Number of factors (for PLS), number of hidden neurons (for ANN), and optimum mtry, i.e., number of variables used for searching the best split at each node (for RF)
– Indicates that the model never reached the 100% correct classification for the Class 2 validation-set individuals at any range of prediction values.