Table 6 Reports for each method the average score obtained during Grid Search and values of hyperparameters most frequently selected during k-fold nested-cross-validation
From: A multi-expert ensemble system for predicting Alzheimer transition using clinical features
| Â | Classifier models | Average score Accuracy on validation set (%) | Best hyperparameters |
|---|---|---|---|
Ensemble proposed | NN | 78.5 | Optimizer: Adam |
Batch size: 60 | |||
Epochs: 100 | |||
Number of hidden units: 32 | |||
RF | 81.7 | Max depth: 80 | |
Min samples for leaf: 3 | |||
Min samples for split: 12 | |||
Number of estimators: 100 | |||
SVM | 83.5 | C: 0.1; | |
\(\gamma\): 1; | |||
Kernel: radial basic function | |||
| Â | AdaBoost | 80.6 | Algorithm: SAMME |
| Â | Learning rate: 0.1 | ||
| Â | Number of estimators: 250 | ||
| Â | MLP | 72.3 | Activation: identity |
| Â | Batch size: 20 | ||
| Â | Epochs: 80 | ||
| Â | Optimizer: Adam | ||
| Â | Number of hidden units: 16 | ||
|  | NB | – | – |
| Â | DT | 78.3 | Criterion of split: Gini |
| Â | Max depth: 2 | ||
| Â | Min samples for leaf: 5 | ||
| Â | Split method: best | ||
| Â | KNN | 54.2 | Distance metric: manhattan |
| Â | Number of neighbors: 19 | ||
| Â | LR | 78.6 | C: 0.0885 |
| Â | Penalty: L1 | ||
| Â | Solver: Newton-cg |