A random forest guided tour

This article is a guided review of the random forest algorithm, introduced by Leo Breiman in 2001, which has become an extremely successful general-purpose method for both classification and regression. The approach builds several randomized decision trees and aggregates their predictions by averaging. The authors highlight why the method is so widely used: it performs excellently in settings where the number of variables is much larger than the number of observations, is versatile enough to be adapted to many ad hoc learning tasks, scales to large problems, and returns useful measures of variable importance.

The review then synthesizes the most recent theoretical and methodological developments for random forests, placing emphasis on the mathematical forces that drive the algorithm. Particular attention is given to the selection of parameters, the resampling mechanism underlying the trees, and variable importance measures. Written to give non-experts easy access to the main ideas, the paper serves as an accessible bridge between random forests' strong empirical track record and the growing body of theory explaining why they work.