Open Access
Table 2
Optimized hyperparameter values for the XGBoost model.
| Hyperparameter | Value | Description |
|---|---|---|
| colsample_bytree | 0.86 | Fraction of features (columns) used per tree. Helps prevent overfitting by limiting how many features each tree can rely on. |
| gamma | 3.3 | Minimum loss reduction required to make a further partition on a leaf node. Higher values help control complexity by preventing excessive splitting. |
| learning_rate | 0.09 | Shrinkage factor that reduces each tree’s contribution. Lower rates often improve accuracy but require more trees. |
| max_depth | 7 | Maximum number of levels in each tree. Deeper trees can learn complex patterns but risk overfitting. |
| min_child_weight | 7 | Minimum sum of instance weights needed in a child (leaf). Larger values restrict the model from creating leaves with few samples. |
| n_estimators | 291 | Number of boosting rounds (trees). More rounds can improve performance but also increase overfitting risk. |
| scale_pos_weight | 208.8 | Balance for class imbalance. Higher values make the model pay more attention to the minority (positive) class. |
| subsample | 0.8 | Fraction of training data sampled per tree. Lower fractions help reduce correlation between trees but can skip important examples. |
Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.
Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.
Initial download of the metrics may take a while.