AI GitHub
XGBoost for gradient-boosted trees
Train boosted tree models for tasks such as tabular classification, regression and ranking.
Key Takeaways
- XGBoost builds trees that incrementally improve an objective and exposes regularization and sampling settings.
- Choose the objective and data split to match the application.
- Start with a modest tree depth and evaluate on a held-out split.
- High validation scores can reflect leakage from identifiers or time.
How it works
XGBoost builds trees that incrementally improve an objective and exposes regularization and sampling settings. Interfaces vary by language but share the underlying modeling approach.
What you need
Choose the objective and data split to match the application. Check which feature encodings and missing-value behaviors are needed by both training and serving.
What to try first
Start with a modest tree depth and evaluate on a held-out split. Use early stopping and compare against a simpler baseline with the same features.
Compare calibration as well as classification accuracy if the application uses predicted probabilities. A better ranker can still produce probability estimates that require adjustment.
Where to look for failures
High validation scores can reflect leakage from identifiers or time. Review feature construction and check the scoring metric before tuning many parameters.
Related Franklin pages
Keep data splits, features and metric settings fixed when comparing tabular baselines. Measure model size and serving time alongside predictive performance.
- Train gradient-boosted trees using histogram-based splitting and configurable leaf growth. LightGBM for efficient tree boosting
- Train and compare candidate models using documented presets and resource budgets. AutoGluon for tabular and multimodal baselines
For a primer on learning from data and evaluating predictions, see Franklin’s glossary. Machine learning
Project documentation
Try it for your workflow
Explore the project’s documentation for setup and usage, or save this page to a collection for later.
Read the project documentation ↗