Bootstrap aggregating (also termed as bagging) is used to create thousands of decision trees that are adequately different from each other. To ensure minimal correlation between trees, each tree is generated from a random subset of the training data, using a random subset of predictor variables.
