Random forests: An ensemble of the classification or the regression trees (see the classification and regression tree methods) that have been fitted to the same n observations, but with the random weights attained by use of the bootstrap. In addition randomness is supplied by selecting a small fraction of covariates for split point determination only in each inner node of these trees.

The final predictions are then attained by averaging the predictions attained from each tree in the forest. The empirical and theoretical investigations have shown that such an aggregation over the multiple tree-structured models helps to get better upon the prediction accuracy of single trees.