Rule quality measures for rule induction systems: Description and evaluation

A rule quality measure is important to a rule induction system for determin ing when to stop generalization or specialization. Such measures are also i mportant to a rule-based classification procedure for resolving conflicts a mong rules. We describe a number of statistical and empirical rule quality formulas and present an experimental comparison of these formulas on a numb er of standard machine learning datasets. We also present a meta-learning m ethod for generating a set of formula-behavior rules from the experimental results which show the relationships between a formula's performance and th e characteristics of a dataset. These formula-behavior rules arc combined i nto formula-selection rules that can be used in a rule induction system to select a rule quality formula before rule induction. We will report the exp erimental results showing the effects of formula-selection on the predictiv e performance of a rule induction system.