Table 5.
VSURF classification performance. Variable selection using random forests (VSURF) classification performance based on bands selected at prediction and interpretation steps (Table 3) for the experiment containing three treatments (control, mock-inoculated, and inoculated seedlings).
| Comparison | Model | Data set | Accuracy | Proportion correctly classified | ||
|---|---|---|---|---|---|---|
| Control | Mock-inoculated | Inoculated | ||||
| All groups | Prediction | Training | 1.000 | 1.000 | 1.000 | 1.000 |
| Testing | 0.562 | 0.515 | 0.457 | 0.703 | ||
| Interpretation | Training | 1.000 | 1.000 | 1.000 | 1.000 | |
| Testing | 0.600 | 0.485 | 0.457 | 0.838 | ||
| Mock vs. inoculated | Prediction | Training | 1.000 | — | 1.000 | 1.000 |
| Testing | 0.792 | — | 0.743 | 0.838 | ||
| Interpretation | Training | 1.000 | — | 1.000 | 1.000 | |
| Testing | 0.806 | — | 0.743 | 0.865 | ||
| Control vs. inoculated | Prediction | Training | 1.000 | 1.000 | — | 1.000 |
| Testing | 0.657 | 0.485 | — | 0.811 | ||
| Interpretation | Training | 1.000 | 1.000 | — | 1.000 | |
| Testing | 0.657 | 0.485 | — | 0.811 | ||