# Pre-pruning decision trees

When building a decision tree to avoid overfitting you may want to stop the branching process even if it could increase the accuracy on the training data . One method to do this is to put in requirements on the final form of the decision tree before you start training. The requirements might be:

• Max depth of the decision tree in any one branch.
• Minimum size of a leaf node for training data.
• Minimum size of a parent node for training data.