XGBoost

XGBoost (eXtreme Gradient Boosting) is an optimized, open-source library providing a scalable, distributed implementation of gradient boosted decision trees.

This is the engine for high-performance predictive modeling: an optimized implementation of the Gradient Boosting framework. XGBoost is engineered for efficiency, often running over 10x faster than other frameworks on a single machine, and handles billions of examples in distributed environments (Spark, Dask). It uses decision trees as base learners, incorporating regularization (L1/L2) to actively combat overfitting, a key differentiator from traditional Gradient Boosting Machines (GBM). The library is the de facto standard for structured data problems, consistently dominating Kaggle competitions and delivering state-of-the-art results across classification, regression, and ranking tasks in languages like Python, R, and Java.

https://xgboost.ai