Table 3 Quality parameters for spectral classification based on different sites. Separation by PCA-LDA was relatively poor. However, SVM performed much better, indicating that knotweed leaf samples can be differentiated by the site at which they were collected using this method
Algorithm | Class | Accuracy | Sensitivity | Specificity |
|---|---|---|---|---|
PCA-LDA (10 PCs, 91% explained variance) | SRC | 85% | 80% | 90% |
SLM | 70% | 57% | 83% | |
SOM | 60% | 24% | 97% | |
SAP | 62% | 35% | 89% | |
ESA | 73% | 52% | 95% | |
ESB | 56% | 20% | 92% | |
EDB | 65% | 43% | 88% | |
SVM | SRC | 100% | 100% | 100% |
SLM | 100% | 100% | 100% | |
SOM | 99% | 98% | 100% | |
SAP | 100% | 100% | 100% | |
ESA | 100% | 100% | 100% | |
ESB | 97% | 94% | 100% | |
EDB | 99% | 100% | 99% |