Chunk #64 — The Methods — How Do Ensemble Methods Work? — Random Forests

Source

An introduction to recursive partitioning: rationale, application, and characteristics of classification and regression trees, bagging, and random forests.

Embedded

yes

Text

In addition to the smoothing of hard decision boundaries, the random selection of splitting variables in random forests allows predictor variables, that were otherwise outplayed by a stronger competitor, to enter the ensemble: If the stronger competitor cannot be selected, a new variable has a chance to be included in the model – and may reveal interaction effects with other variables that otherwise would have been missed.