Table 2 Hyperparameters achieved for the machine learning models after grid search and pipelines. We can see the models, the evaluated parameters chosen for both two classes and multiclass, and their descriptions
Algorithms | Hyperparameter | Two-class | Multi-class | Description |
|---|---|---|---|---|
DTC | criterion | entropy | gini | Evaluates the quality of a division |
max_depth | None | N/A | Maximum depth of the tree | |
min_samples_leaf | 4 | 2 | Minimum number of samples required per leaf node | |
min_samples_split | 10 | 10 | Samples needed to split an internal node | |
MLP | activation | logistic | relu | Function that activates the hidden layer |
hidden_layer_sizes | (100,) | (100,50) | Number of neurons of the i-th hidden layer | |
learning_rate | constant | constant | Learning rate programming for weight updates | |
solver | Adam | Adam | Solver for weight optimization | |
KNN | algorithm | auto | auto | Calculate nearest neighbors |
leaf_size | 1 | 1 | Leaf size passed to BallTree or KDTree | |
n_neighbors | 1 | 1 | Number of neighbors | |
p | 2 | 1 | Indicates the power for the Minkowski metric | |
weights | ‘uniform’ | ‘uniform’ | Used in prediction | |
SGDC | alpha | 0.001 | 0.01 | Constant that multiplies the regularization term |
loss | ‘hinge’ | ‘hinge’ | ||
max_iter | 2000 | 1000 | Number of epochs performed on training data | |
penalty | ‘l2’ | ‘l1’ | Regularization term | |
weights | ‘uniform’ | ‘uniform’ | Used in prediction | |
ETC | min_samples_split | 4 | N/A | Samples needed to split an internal node |
n_estimators | 150 | 300 | Number of trees in the forest | |
random_state | 20 | 20 | Controls the bootstrapping of the samples | |
weights | ‘uniform’ | ‘uniform’ | Used in prediction | |
SVM | C | 10 | 10 | Regularization parameter |
gamma | ‘scale’ | ‘auto’ | Is a coefficient | |
kernel | ‘rbf’ | ‘rbf’ | Is the type of kernel in use | |
RFC | max_depth | None | None | Maximum depth of the tree |
min_samples_split | 2 | 2 | Samples needed to split an internal node | |
n_estimators | 500 | 500 | Number of trees in the forest | |
random_state | 40 | 40 | Controls the bootstrapping of the samples | |
GB | learning_rate | 0.1 | 0.1 | Is a compensation between n_estimators and learning_rate |
max_depth | 5 | 7 | Maximum depth of the regression estimators | |
n_estimators | 200 | 200 | The number of stages to be performed | |
random_state | 40 | 10 | At each iteration of reinforcement controls the seed |