Table 1.
Prediction of testing data sets. S (susceptible) and R (resistant/tolerant) refer to reference dataset. SVM predict 10 out of 11 samples, being the less accurate model. The training set model accuracy was calculated fractionating the number of correct predictions (true positive + true negative) by the total number (true positive + true negative + false positive + false negative) and using the sparsity values for the proportion of features used in training each model.
| Reference Data | S | R | S | R | S | S | R | S | S | R | R | Model Accuracy |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| SVM | S | R | S | R | S | S | R | R | S | R | R | 0.89 |
| NSC | S | R | S | R | S | S | R | S | S | R | R | 0.96 |
| PLDA | S | R | S | R | S | S | R | S | S | R | R | 0.93 |
| PLDA2 | S | R | S | R | S | S | R | S | S | R | R | 0.96 |
| VoomDLDA | S | R | S | R | S | S | R | S | S | R | R | 0.95 |
| VoomNSC | S | R | S | R | S | S | R | S | S | R | R | 0.96 |
| VoomNBLDA | S | R | S | R | S | S | R | S | S | R | R | 0.95 |
SVM = Support Vector Machine; NSC = Supervised Normalized Cut; PLDA = Parallel Latent Dirichlet Allocation; Voom = Variance modeling at the observational level; DLDA and NBLDA are diagonal discriminant classifiers.