Meta-Pattern Extraction: Mining Cycles

Inductive computing, comprised of machine learaing, data mining, and knowledge discovery, seeks to extract causal patterns from data. Meta-patterns are patterns of extrax:ted patterns. The meta-pattern we study here is the cycle. In this paper, we illustrate the ubiquity of cycles anti argue that as responsible knowledge engineers, we need to identify and actively seek cycles out of data. We present a methodology for cycle mining and exemplify such a pursuit in a generic relation’,d model as well as briefly discussing the authors’implementation, computer system INDED. System INDED is a machine learning system implementing inductive logic programming techniques of pattern extraction. Introduction and Motivation Knowledge discovery in databases has been defined as the sou-trivial process of identifying valid, novel, potentially useful, and understandable patterns in data [5]. A pattern is often denoted in the form of an IF-THEN rule (IF antecedent THEN consequenQ, where the antecedent and consequent are logical conjunctions of predicates (first order logic) propositions (propositional logic) [10]. In [9], the author observes that knowledge can take on more coiuplex forms than a simple implication as a causal chain or network by interconnecting the consequent of one rule to the antecedent of another. Acyclic graphs are used extensively as knowledge representation constructs in knowledge discovery in databases *C,c~pyright gt.’ 1999, American Association for Artificial InteUigence (www.aaaJ.org}. All rights reserved. t Partially supported under Grant 9806184 by the National Science Foundation. as seen in [6] and [10]. It is largely our goal to bring to the reader’s attention the benefit of employing a graph capable of possessing cycles to represent learned knowledge and identifying the appropriate propositional vertices as cycle participants. Why Cycles Typically Exist in Data Much of our social construction of reality stems directly from the physical cycles of the earth revolving around the sun, and its rotation on its axis: the temporal calendar by which we live. We invent and self-impose other cycles based on these: the seasons and holidays of the year, the organization of the academic and fiscal years. These cycles arc the basis for much of our common sense, domain kzmwledge, and they spawn related cycles such as the cycle of seasons justifying the cycle of weather events we might experience. Human behavior is based on reinforcement, repetition, and routine [7]. This human behavior manifests in the telephone calls and purchases we make, the food we eat, and the hardships, such as illness, we experience. In our speech, we often allude to a negative sequence of human behaviors, such as those manifested as overall poor quality in manufacturing, or more profoundly, violence in our society, as "being caught in a vicious cycle." Or we embrace a positive sequence of behaviors and try to secure a cycle by reinforcing excellent employee performance or by instilling good habits in our children. Cycles appear in our creative pursuits, as well. An analysis of the tonality of virtually any musical work of Bach will indicate a traversal around the circle of fifths [3]. A study of Escher’s art will haw~ the observer entranced in subtle, circular, repetitious patterns [8]. The authors contend that evidence of this cyclic nature of hunmns and our world exists in our data. 466 SELTZER From: Proceedings of the Twelfth International FLAIRS Conference. Copyright © 1999, AAAI (www.aaai.org). All rights reserved.