Database Reference
In-Depth Information
Chapter 18
Applications of Frequent Pattern Mining
Charu C. Aggarwal
Abstract
Frequent pattern mining has broad applications which encompass cluster-
ing, classification, software bug detection, recommendations, and a wide variety of
other problems. In fact, the greatest utility of frequent pattern mining (unlike other
major data mining problems such as outlier analysis and classification), is as an
intermediate tool
to provide pattern-centered insights for a variety of problems. In
this chapter, we will study a wide variety of applications of frequent pattern mining.
The purpose of this chapter is not to provide a detailed description of every possible
application, but to provide the reader an overview of what is possible with the use of
methods such as frequent pattern mining.
Keywords
Frequent pattern mining
1
Introduction
This chapter provides an overview of the key applications of frequent pattern mining.
Frequent pattern mining was first proposed by Agrawal et al in 1993 [
11
,
13
]. Since
then, a wide variety of tree-based and pattern growth-based algorithms have been
proposed for the problem [
16
,
27
,
59
,
114
,
141
]. An overview of algorithms for
frequent pattern mining may be found in [
60
].
Frequent pattern mining is one of the unusual problems in data mining, where
the size of the output may sometimes be comparable or larger than the size of the
input. For example, in the context of a database with a few thousand transactions, it
is often possible to obtain a number of frequent patterns which are of the same or
significantly larger magnitude. Therefore, a question arises as to the utility of such
large outputs from a mining algorithm, if they do not provide a
concise
summary or
characterization of the underlying data. In practice, these outputs are often used as
intermediary steps
in other data mining applications. Therefore, the greatest utility
of frequent pattern mining algorithms is as an intermediary step, rather than as a
goal in of itself. Without using some form of post-processing, it is often difficult to
