SEARCH

SEARCH BY CITATION

Abstract

Clustering methods are one of the most popular approaches to data mining. They have been successfully used in virtually any field covering domains such as economics, marketing, bioinformatics, engineering, and many others. The classic cluster algorithms require static data structures. However, there is an increasing need to address changing data patterns. On the one hand, this need is generated by the rapidly growing amount of data that is collected by modern information systems and that has led to an increasing interest in data mining as its whole again. On the other hand, modern economies and markets do not deal with stable settings any longer but are facing the challenge to adapt to constantly changing environments. These include seasonal changes but also long-term trends that structurally change whole economies, wipe out companies that cannot adapt to these trends, and create opportunities for entrepreneurs who establish large multinational corporations virtually out of nothing in just one decade or two. Hence, it is essential for almost any organization to address these changes. Obviously, players that have information on changes first possibly obtain a strategic advantage over their competitors. This has motivated an increasing number of researchers to enrich and extend classic static clustering algorithms by dynamic derivatives. In the past decades, very promising approaches have been suggested; some selected ones will be introduced in this review. © 2012 Wiley Periodicals, Inc.