Gradient Boosting

#Terms

WHAT IS… ?

Gradient boosting is an ensemble learning method in which a set of weak prediction models are iteratively made stronger by aggregating them to produce a stronger base model, which is then used to make a new set of even stronger models, and so on. Gradient boosting has been shown to work very well on many practical data mining tasks.

HOW IS IT USED ?

The simple prediction models are weak on their own but when combined they form a powerful base model. Gradient boosting also improves the predictive performance as new samples become available.