Table 3.
Performance comparison of different ML and NN models for different types of error (e1,e2,e3)
| (e1,e2,e3) | SVM | GB | RF | MNB | LR1 | LR2 | MLP | CNN |
|---|---|---|---|---|---|---|---|---|
| F1-micro | ||||||||
| (0.5, 0.1, 0.4) | 0.96 | 0.79 | 0.98 | 0.98 | 0.30 | 0.98 | 0.98 | 0.75 |
| (0.5, 0.4, 0.1) | 0.99 | 0.82 | 1.00 | 1.00 | 0.43 | 1.00 | 1.00 | 0.81 |
| (0.3, 0.1, 0.4) | 0.98 | 0.87 | 0.98 | 0.99 | 0.54 | 0.99 | 0.99 | 0.74 |
| (0.0, 0.7, 0.2) | 0.99 | 0.83 | 1.00 | 1.00 | 0.66 | 1.00 | 1.00 | 0.86 |
| (0.0, 0.2, 0.7) | 0.89 | 0.58 | 0.81 | 0.91 | 0.51 | 0.87 | 0.91 | 0.59 |
We consider several existing supervised ML methods, as well as NN models (i.e., MLP and CNN). For each experiment, we use 10-fold cross-validation. We use F1-micro to quantify the performance as defined in Classification performance metrics. Bold values represent the best results