How Useful Is Relevance
Rich Caruana, Dayne Freitag · 1994
Eliminating irrelevant attributes prior to induction boosts the performance of many learning algorithms. Relevance, however, is no guarantee of usefulness to a particular learner. We test two methods of finding relevant attributes, FOCUS and RELIEF, to see how the attributes they select perform with ID3/C4.5 on two learning problems from a calendar scheduling domain. A more direct attribute selection procedure, hillclimbing in attribute space, finds superior attribute sets. 1 INTRODUCTION An attribute that is irrelevant is not useful for induction. But not all attributes that are relevant are necessarily useful for induction either. Several methods for estimating attribute relevance have been devised. Are the attributes selected by these methods good attributes to use for learning? Let A be the set of attributes potentially available to a learner L trying to learn problem P given typical training sets. Imagine that the learner is given only the attributes in A ae A. We call some attri...