Table 2.
Species with 90% or more of specimens predicted to correct species if no threshold was set
| Species | Number of specimens | No threshold | Threshold 0.5 | |||
|---|---|---|---|---|---|---|
| True positive (%) | False positive | True positive (%) | Not species level | False positive | ||
| Cylindera germanica | 39 | 100 | 12 | 94.9 | 2 | 4 |
| Cychrus caraboides | 57 | 98.2 | 13 | 96.5 | 2 | 4 |
| Calosoma inquisitor | 39 | 97.4 | 8 | 97.4 | 0 | 3 |
| Carabus glabratus | 27 | 96.3 | 2 | 88.9 | 2 | NA |
| Lebia chlorocephala | 67 | 95.5 | 3 | 94.0 | 1 | 3 |
| Broscus cephalotes | 111 | 93.7 | 14 | 92.8 | 3 | 4 |
| Cicindela campestris | 88 | 93.2 | NA | 92.0 | 3 | NA |
| Nebria complanata | 55 | 92.7 | 1 | 89.1 | 4 | NA |
| Carabus problematicus | 107 | 91.6 | 16 | 87.9 | 7 | 14 |
| Chlaenius nigricornis | 40 | 90.0 | 4 | 82.5 | 5 | 3 |
Number of specimens in test dataset, percentage of specimens predicted to the correct species, and false positives (i.e., the number of specimens predicted to the wrong species). For threshold 0.5, the percentage of specimens in species predicted to the correct species, the number of specimens that did not meet the threshold, thus not predicted on species level, and false positives (i.e., the number of specimens predicted to the wrong species).