Combining Bagging and Random Subspaces to Create Better Ensembles

Random forests are one of the best performing methods for constructing ensembles. They derive their strength from two aspects: using random subsamples of the training data (as in bagging) and random