Boosting with Multi-Way Branching in Decision Trees

Part of Advances in Neural Information Processing Systems 12 (NIPS 1999)

Bibtex Metadata Paper

Authors

Yishay Mansour, David McAllester

Abstract

It is known that decision tree learning can be viewed as a form of boosting. However, existing boosting theorems for decision tree learning allow only binary-branching trees and the generalization to multi-branching trees is not immediate. Practical decision tree al(cid:173) gorithms, such as CART and C4.5, implement a trade-off between the number of branches and the improvement in tree quality as measured by an index function. Here we give a boosting justifica(cid:173) tion for a particular quantitative trade-off curve. Our main theorem states, in essence, that if we require an improvement proportional to the log of the number of branches then top-down greedy con(cid:173) struction of decision trees remains an effective boosting algorithm.