Concept dispersion, feature interaction and their effect on particular sources of bias in machine learning

Many fast, efficient induction algorithms have been produced which perform well across many domains. However, in more difficult problems - where low-l evel representations are present - these algorithms can perform poorly and abstract feature construction is often required. Greedy hill-climbing algor ithms are harmed by feature interaction, where one attribute alone provides little information about the concept. We argue that the need for, and effe cts of feature construction are often swamped by the bias of the base learn er. Other methods of class formation which are not as susceptable to the pr oblems associated with low-level representation should be explored and augm ented as a basis for feature construction. We present an alternative, infor mation theoretic based approach to detecting feature interaction and develo p harsher constraints on hypothesis generation, based on relative and absol ute measures. (C) 1998 Elsevier Science B.V. All rights reserved.