Constraint-Based Pattern Mining - Frequent Pattern Mining - page 144

Chapter 7
Constraint-Based Pattern Mining

Siegfried Nijssen and Albrecht Zimmermann

Abstract Many pattern mining systems are designed to solve one specific problem,
such as frequent, closed or maximal frequent itemset mining, efficiently. Even though
efficient, their specialized nature can make these systems difficult to apply in other
situations than the one they were designed for. This chapter provides an overview of
generic constraint-based mining systems. Constraint-based pattern mining systems
are systems that with minimal effort can be programmed to find different types of pat-
terns satisfying constraints. They achieve this genericity by providing (1) high-level
languages in which programmers can easily specify constraints; (2) generic search
algorithms that find patterns for any task expressed in the specification language.
The development of generic systems requires an understanding of different classes
of constraints. This chapter will first provide an overview of such classes constraints,
followed by a discussion of search algorithms and specification languages.

Keywords Constraints · Languages · Inductive databases · Search algorithms

1 Introduction

A key component of a pattern mining system is the constraint that is used by the
system. A frequent itemset mining system, for instance, is characterized by the use
of a minimum support constraint; an association rule mining system, similary, is
identified by a minimum confidence constraint. Constraints define to a large degree
which task a pattern mining system is performing.
However, the focus of many pattern mining systems on one particular type of
constraint can make their use cumbersome. As an example, consider a frequent
itemset mining system that one wishes to apply in a context where the utility of the