Table 3.
Model parameters.
| Parameter | Value |
| XGBoost | |
| Base Learner | Gradient boosted tree |
| Tree construction algorithm | Exact greedy |
| Learning rate | 0.0991 |
| Lagrange multiplier | 0 |
| Number of gradient boosted trees | 80 |
| Maximum depth of a tree | 6 |
| Minimum sum of instance weight | 1 |
| Subsample ratio of the training instances | 1 |
| Sampling method | Uniform |
| L2 regularization term on weights | 1 |
| Tree growing policy | Depthwise |
| Evaluation metrics for validation data | Negative log likelihood |
| SVM | |
| Kernel | Linear |
| Degree of the polynomial kernel | 3 |
| Kernel coefficient | Scale |
| Maximum iterations | No constraint |
| Shrinking heuristic | True |
| Probability estimates | False |
| Tolerance for stopping criterion | 1 × 10−3 |
| Random forest | |
| Number of trees in the forest | 10 |
| Quality of split measure function | Entropy |
| Minimum number of samples to split | 2 |
| Minimum number of samples at a leaf node | 1 |
| Use bootstrap samples for building trees | True |
| Number of jobs to run in parallel | 1 |
| CatBoost | |
| Number of boosting rounds | 20 |
| Learning rate | 0.44 |
| Maximum depth of a tree | 5 |
| Maximum number of trees | 1000 |
| Random seed | 0 |
| Sample weight frequency | Per tree level |
| Tree growing policy | Symmetric tree |
| Maximum number of leaves | 31 |
| LightGBM | |
| Number of decision trees | 20 |
| Bagging fraction | 1 |
| Number of threads in the physical core | 8 |
| Maximum depth of a tree | 6 |
| Number of boosting iterations | 100 |
| Learning rate | 0.1 |
| Maximum number of leaves on one tree | Serial |
| Bagging random seed | 3 |
| Dropout rate | 0.1 |