Decision Tree Pruning

Method to avoid Overfitting for Decision Tree when there are too many branches that might even reflect Outlier or noise.

Pruned trees are smaller, faster, less complex and better at classifying unseen data.

Prepruning

Halt Decision Tree Induction early if goodness measure falls below a threshold.

Postpruning

Progressively remove terminal nodes from an overfit tree and use testset or Information Gain to decide which is the best pruned tree.