Classifier | Hyper-parameter name | Description | Optimized value space |
MLP | hidden_layer_sizes | Number of hidden layers and neurons | [100, 70, 50] |
Activation | Output function of non-input neuron | [logistic, relu, tanh] | |
learning_rate | Regularization parameter | [0.1, 0.001, 0.0001] | |
Alpha | Controls step size during weight updates | [0.0001, 0.01, 0.1] | |
Solver | Weight optimization method | [lbfgs, sgd, adam] | |
RF ET | n_estimators | Number of trees in the forest | [70, 100, 120] |
max_depth | Maximum depth of the tree. | [3, 5, 8] | |
min_samples_leaf | Minimum number of samples at a leaf node. | RF = [2, 5, 10], ET = [1, 2, 5] | |
Criterion | Measure the quality of a split | [gini, entropy] | |
GBM | n_estimators | Number of tress. | [70, 100, 120] |
learning_rate | Shrinkage factor for each tree | [0.1, 0.001] | |
min_samples_leaf | Minimum number of samples needed to be at a leaf node | [5, 10] | |
max_depth | Maximum depth of individual tree | [3, 5, 7] | |
Loss | The loss function in the boosting process to be optimized. | [log_loss, exponential] | |
Adaboost | n_estimators | Number of trees. | [70, 100, 120] |
learning_rate | Boosting learning rate | [0.1, 0.001] | |
XGBoost | n_estimators | Number of trees. | [70, 100, 120] |
learning_rate | Shrinkage factor of each tree | [0.1, 0.001] | |
max_depth | Tree depth | [5, 10, 20] | |
subsample | Subsample ratio of training samples | [0.1, 0.5, 1] | |
LightGBM | n_estimators | Number of gradient boosted trees. | [70, 100, 120] |
learning_rate | Shrinkage factor of each tree | [0.1, 0.001] | |
max_depth | Maximum tree depth for base learners. | [10, 20, 30] | |
num_leaves | Number of leaves for each tree | [5, 8, 15] |