The general method for reining in model complexity to avoid overfitting is called model regularization. Techniques include tree pruning (cutting a classification tree back when it has become too large), feature selection, and employing explicit complexity penalties into the objective function used for modeling.
