Table 3.
The hyperparameters tuning of models.
| Parameters | Definition | LightGBM | XGBoost | CatBoost | AdaBoost | SVR | MLP |
|---|---|---|---|---|---|---|---|
| Boosting_type | Boosting method | gbdt | – | – | – | – | – |
| Learning_rate | Boosting learning rate | 0.01 | 0.1 | 0.01 | 0.001 | – | – |
| Max_depth | Maximum tree depth for base learners | 5 | 3 | 7 | – | – | – |
| Min_child_samples | Minimum number of data needed in a child | 100 | – | – | – | – | – |
| Num_leaves | Maximum tree leaves for base learners | 30 | – | – | – | – | – |
| Subsample | Subsample ratio of the training instance | 0.1 | 0.5 | 0.7 | – | – | – |
| Booster | Type of booster | gbtree | – | – | – | – | – |
| Gamma | Minimum loss reduction | – | – | – | – | – | – |
| n_estimator | Number of gradient-boosted trees | – | – | 500 | 100 | – | – |
| Min_child_weight | The minimum sum of instance weight(hessian) | – | 5 | – | – | – | – |
| Loss | Loss function | – | – | – | Linear | – | – |
| Kernel | Kernel type | – | – | – | – | rbf | – |
| C | Regularization parameter | – | – | – | – | 10 | – |
| Gamma | kernel parameter | – | – | – | – | 0.08 | – |
| Epsilon | The epsilon-tube | – | – | – | – | 0.01 | – |
| Hidden_layer_sizes | Number of neurons in the ith hidden layer | – | – | – | – | – | 20 |
| Max_iter | The maximum number of iterations | – | – | – | – | – | 20 |
| Learning_rate_init | The initial learning rate | – | – | – | – | – | 0.08 |
| Learning_rate | Learning rate | – | – | – | – | – | Invscaling |
| Activation | Activation function | – | – | – | – | – | relu |