Irrelevant Features and the Subset Selection Problem-英文文献.pdf

Published in Willia m W Cohen Haym Hirsh eds Machine Learning Proceedings of the Eleventh International Conference Morgan Kaufmann Publishers San Francisco CA Irrelevant Features and the Subset Selection Problem George H John Ron Kohavi Karl Peger Computer Science Dept Computer Science Dept Computer Science Dept Stanford University Stanford University Stanford University Stanford CA Stanford CA Stanford CA gjohnCSStanfordEDU ronnykCSStanfordEDU kpflegerCSStanfordEDU Abstract Correlated We address the problem of nding a subset 0 1 of features that allows a sup ervised induc A1 Irrelevant tion algorithm to induce small highaccuracy concepts We examine notions of relevance 0 1 0 1 and irrelevance and show that the denitions B0 A0 A0 1 used in the machine learning literature do not adequately partition the features into useful 0 1 0 1 0 1 categories of relevance We present deni 0 B1 B1 1 0 A1 tions for irrelevance and for two degrees of relevance These denitions improve our un 0 1 0 1 0 1 derstanding of the b ehavior of previous sub 0 1 0 B0 B0 1 set selection algorithms and help de