← AI Terminology
XGBoost
XGBoost is a highly optimised gradient-boosting library for decision tree ensembles, famous for dominating tabular ML competitions and industry scorecards.
It popularised scalable, regularised boosting in production.
It popularised scalable, regularised boosting in production.
Why It Matters in AI
For structured/tabular data, gradient-boosted trees often still beat deep nets. XGBoost became synonymous with practical ML winning solutions and remains a default enterprise tool alongside LightGBM/CatBoost.
Key Points
| Aspect | Description |
|---|---|
| Use | Fraud, ranking, pricing, Kaggle-style problems |
| Type | Gradient boosted decision trees (GBDT) |
| Related | Boosting, random forests |
| Features | Regularisation, sparsity awareness, parallel training |
| Strengths | Tabular accuracy, missing values, feature importance |
| Competitors | LightGBM, CatBoost, HistGradientBoosting |
Simple Analogy
A committee of simple yes/no tree experts added one by one, each correcting the previous committee’s mistakes — engineered for speed.
Common Usage Examples
xgboost.XGBClassifier- Rank:XGB for learning-to-rank
- SHAP on XGBoost for explainability
- Still try GBDT before tabular DL
Summary
In short: XGBoost is the iconic gradient-boosted trees library — still a top choice for high-accuracy tabular machine learning.