Nature-inspired clustering

Research output: Chapter in Book/Report/Conference proceedingChapterpeer-review

Abstract

We survey data clustering techniques that are inspired by natural systems. It is instructive to divide this group into two distinct classes: those methods derived directly from clustering phenomena in nature, such as the self-organizing behavior of some ants, or the schooling of fish and flocking of birds; and those methods that view clustering as an optimization problem, and apply nature-inspired general heuristics (or meta-heuristics) to this problem. The first group work largely in a bottom-up fashion, making them potentially powerful for parallelization and for operating in dynamic and noisy clustering applications, such as distributed robotics. The second group has the advantage that the clustering problem can be specified very precisely using one or more mathematical objective functions; then, due to the general-purpose properties of the heuristics, very good performance against these objectives can be achieved. This chapter charts the progress in these two broad classes of nature-inspired clustering algorithms and points out some further prospects in this area.

Original languageEnglish
Title of host publicationHandbook of Cluster Analysis
EditorsChristian Hennig, Marina Meila, Fionn Murtagh, Roberto Rocci
PublisherCRC Press and Balkema
Pages419-440
Number of pages22
ISBN (Electronic)9781466551893
ISBN (Print)9781466551886
DOIs
Publication statusPublished - 1 Jan 2015

Fingerprint

Dive into the research topics of 'Nature-inspired clustering'. Together they form a unique fingerprint.

Cite this