Pushing constraints to detect local patterns

Abstract

The main position of this paper is that constraints can be a very useful tool in the search for local patterns. The justification for our position is twofold. On one hand, pushing constraints makes feasible the computation of frequent patterns at very low frequency levels, which is where local patterns are. On the other hand constraints can be exploited to guide the search for those patterns showing deviating, surprising characteristics. We first review the many definitions of local patterns. This review leads us to justify our position. We then provide a survey of techniques for pushing constraint into the frequent pattern computation. © Springer-Verlag Berlin Heidelberg 2005.