Title: Impact of subsampling and pruning on random forests

Abstract: Random forests are ensemble learning methods introduced by Breiman (2001) that operate by averaging several decision trees built on a randomly selected subspace of the data set. Despite their widespread use in practice, the respective roles of the different mechanisms at work in Breiman's forests are not yet fully understood, neither is the tuning of the corresponding parameters. In this paper, we study the influence of two parameters , namely the subsampling rate and the tree depth, on Breiman's forests performance. More precisely, we show that fully developed sub-sampled forests and pruned (without subsampling) forests have similar performances, as long as respective parameters are well chosen. Moreover , experiments show that a proper tuning of subsampling or pruning lead in most cases to an improvement of Breiman's original forests errors.