Decision trees are a suitable family of elementary algorithms for bagging since they are quite complicated and can ultimately achieve zero errors on any training set. We can use a method that uses random subspaces (such as bagging) to reduce the correlation between trees and avoid overfitting. The elementary algorithms are trained on different subsets of the feature space, which are also randomly selected. An ensemble of decision tree models using the random subspace method can be constructed using the following algorithm.
Where the number of objects for training is N and the number of features is 
, proceed as ...
