Table 1 Cohen’s kappa statistic for intra-rater reliability against ground truth at 500 and 20k samples.
From: Molecular function recognition by supervised projection pursuit machine learning
| Â | PCA | PLS | DR SPLOC | |
|---|---|---|---|---|
500 samples—Cohen’s kappa | ||||
3 modes | max | 0.200 | 0.385 | 0.882 |
avg. | 0.181 | 0.247 | 0.835 | |
min | 0.161 | 0.125 | 0.600 | |
8 modes | max | 0.600 | 0.600 | .882 |
avg. | 0.394 | 0.285 | 0.774 | |
min | 0.143 | 0.125 | 0.600 | |
13 modes | max | 0.600 | 0.882 | 0.882 |
avg. | 0.371 | 0.395 | 0.645 | |
min | 0.231 | 0.125 | 0.333 | |
20k samples—Cohen’s kappa | ||||
3 modes | max | 0.143 | 0.565 | 1.000 |
avg. | 0.060 | 0.345 | 0.9825 | |
min | 0.000 | 0.125 | 0.895 | |
8 modes | max | 0.714 | 1.000 | 1.000 |
avg. | 0.549 | 0.806 | 1.000 | |
min | 0.385 | 0.120 | 1.000 | |
13 modes | max | 0.714 | 1.000 | 1.000 |
avg. | 0.602 | 0.822 | 1.000 | |
min | 0.565 | 0.217 | 1.000 | |
