Table 4.
Classification accuracy without feature/operational taxonomic unit (OTU) selection, measured by proportion of correct classifications (PCC)
| Classifier | CBH | CS | CSS | FS | FSH | BP | PDX | PBS | Averages | P values |
|---|---|---|---|---|---|---|---|---|---|---|
|
SVM, Linear C = 1 |
0.920 |
0.911 |
0.583 |
0.940 |
0.598 |
0.354 |
0.468 |
0.695 |
0.684 |
0.022 |
|
SVM, Linear optimized C |
0.920 |
0.911 |
0.622 |
0.980 |
0.585 |
0.383 |
0.485 |
0.709 |
0.699 |
0.038 |
|
SVM, Poly |
0.920 |
0.911 |
0.622 |
0.980 |
0.585 |
0.383 |
0.484 |
0.709 |
0.699 |
0.036 |
|
SVM, RBF |
0.909 |
0.904 |
0.575 |
0.973 |
0.575 |
0.379 |
0.451 |
0.700 |
0.683 |
0.021 |
|
KRR, Poly |
0.913 |
0.918 |
0.581 |
0.954 |
0.598 |
0.377 |
0.482 |
0.709 |
0.692 |
0.027 |
|
KRR, RBF |
0.923 |
0.904 |
0.618 |
0.967 |
0.632 |
0.366 |
0.467 |
0.709 |
0.698 |
0.030 |
|
KNN, K = 1 |
0.496 |
0.360 |
0.195 |
0.451 |
0.305 |
0.249 |
0.419 |
0.291 |
0.346 |
0.002 |
|
KNN, K = 5 |
0.713 |
0.339 |
0.188 |
0.397 |
0.281 |
0.331 |
0.393 |
0.300 |
0.368 |
0.001 |
|
KNN, optimized K |
0.714 |
0.377 |
0.192 |
0.325 |
0.273 |
0.340 |
0.409 |
0.379 |
0.376 |
0.001 |
|
PNN |
0.743 |
0.321 |
0.216 |
0.522 |
0.332 |
0.325 |
0.167 |
0.247 |
0.359 |
0.000 |
|
L2-LR, C = 1 |
0.934 |
0.939 |
0.628 |
0.982 |
0.628 |
0.380 |
0.515 |
0.725 |
0.716 |
0.084* |
|
L2-LR, optimized C |
0.933 |
0.938 |
0.623 |
0.978 |
0.618 |
0.383 |
0.502 |
0.725 |
0.712 |
0.067* |
|
L1-LR, C = 1 |
0.929 |
0.801 |
0.559 |
0.975 |
0.700 |
0.422 |
0.384 |
0.673 |
0.680 |
0.018* |
|
L1-LR, optimized C |
0.928 |
0.903 |
0.561 |
0.981 |
0.690 |
0.445 |
0.412 |
0.692 |
0.702 |
0.039 |
|
RF, default |
0.932 |
0.955 |
0.673 |
0.999 |
0.744 |
0.508 |
0.424 |
0.730 |
0.746 |
0.270* |
|
RF, optimized
|
0.938
|
0.956
|
0.689
|
0.994 |
0.760
|
0.523
|
0.423 |
0.735
|
0.752 |
- |
|
BLR, Laplace priors |
0.927 |
0.927 |
0.634 |
0.962 |
0.622 |
0.387 |
0.452 |
0.727 |
0.705 |
0.042 |
| BLR, Gaussian priors | 0.921 | 0.736 | 0.480 | 0.966 | 0.631 | 0.354 | 0.410 | 0.635 | 0.642 | 0.008 |
The nominally best performing classifier on average over all datasets is marked with bold, and P values of methods whose performance cannot be deemed statistically worse than the nominally best performing method are marked with “*”. The accuracy of the nominally best performing method for each dataset is underlined.