Bottom-Up attribute Selection for the induction of decision Trees

Denis Pomorski (1)

1 LAGIS - Laboratoire d'Automatique, Génie Informatique et Signal

We propose a new methodology, called BUST, for the extraction of joint relationships using induction of decision trees. BUST is the acronym of “Bottom-Up attribute Selection for the induction of decision Trees”: at each node of the tree, a bottom-up approach of attribute selection is used, as opposed to the classical Top-Down approach. This methodology has been developed to solve functional separability problem: the irrelevant or redundant attributes, which should be the last attributes tested in the tree, are rejected.

Domaines: Intelligence artificielle [cs.AI]; Théorie de l'information [cs.IT]

Denis Pomorski. Bottom-Up attribute Selection for the induction of decision Trees. The Multiconference on Computational Engineering in Systems Applications CNRS/IMACS/IEEE-SMC (CESA’2003), Jul 2003, Lille, France. ⟨hal-01509830⟩