Discovering Multi-head Attributional Rules in Large Databases
Date
2001-06
Authors
Głowiński, Cezary
Michalski, Ryszard S.
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
A method for discovering multi-head attributional rules in large databases is presented and illustrated by results from an implemented program. Attributional rules (a.k.a. attributional dependencies) can be viewed as generalizations of standard association rules, because they use more general and expressive conditions than those in the latter ones, and by that can express more concisely inter-attribute relations in a database. Multi-head rules have multiple conditions/statements in their conclusion. The presented method applies AQ learning to create single-head characteristic rules, and then seeks conditions (selectors) that can be transferred to the conclusion part of the rule. Experiments with the program MAR1 (Multi-Head Attributional Rules), implementing the developed method, has produced highly encouraging results.
Description
Keywords
Attributional rules
Citation
(P 01-4) Glowinski, C. and Michalski, R. S., "Discovering Multi-head Attributional Rules in Large Databases," Tenth International Symposium on Intelligent Information Systems, Zakopane, Poland, June, 2001.