Table 1.
Configuration of machine learning pipelines for all folds of the inner NCV loop using the N-AUG and AUG datasets grouped by classifier type. Min Max: min-max normalisation; Standard: standardisation; None: No data scaling; C: regularisation parameter; L1: Lasso penalty term; L2: Ridge penalty term; Elastic Net: Combination of L1 and L2; Criterion: tree-specific splitting criterion; GBTree: tree-based booster; DART: dropout regularized tree-based gradient booster.
| Algorithm | Dataset |
Hyperparameters |
||||
|---|---|---|---|---|---|---|
| Scaler | C | Penalty | Solver | L1 ratio | ||
| LR | N-AUG N-AUG N-AUG |
None None None |
0.19 0.19 0.36 |
L1 L1 Elastic Net |
SAGA SAGA SAGA |
- - 0.85 |
| Scaler | C | Kernel | ||||
| SVM | N-AUG | None | 0.001 | Polynomial | ||
| Scaler | Estimators | Max Depth | Criterion | |||
| RF | N-AUG | None | 40 | 2 | Entropy | |
| Scaler | Estimators | Max Depth | Booster | Gamma | ||
| XGB | AUG | None | 400 | 6 | GBTree | 8.44 |
| AUG | Standard | 360 | 5 | DART | 8.93 | |
| AUG | Standard | 320 | 5 | DART | 8.19 | |
| AUG | Min Max | 240 | 6 | DART | 11.18 | |
| AUG | Min Max | 240 | 6 | DART | 11.18 | |