This Title All WIREs
How to cite this WIREs title:
WIREs Data Mining Knowl Discov
Impact Factor: 7.250

Choosing the number of clusters

Full article on Wiley Online Library:   HTML PDF

Can't access this content? Tell your librarian.

Abstract The issue of determining ‘the right number of clusters’ is attracting ever growing interest. The paper reviews published work on the issue with respect to mixture of distributions, partition, especially in k‐means clustering, and hierarchical cluster structures. Some perspective directions for further developments are outlined. © 2011 John Wiley & Sons, Inc. WIREs Data Mining Knowl Discov 2011 1 252–260 DOI: 10.1002/widm.15 This article is categorized under: Algorithmic Development > Structure Discovery Technologies > Structure Discovery and Clustering

A taxonomy for the approaches to choose the number of clusters as described in the paper. Here DK is domain knowledge in which DA is direct adjustment of algorithm and AP is adjustment by post‐processing of clustering results; GC is modelling of the process of generation of clusters, and CS is cluster structures. This last item is further divided into MD, that is, mixture of distributions; P, partitions; BH, binary hierarchies; and OS, other cluster structures. MD involves HT, which is hypothesis testing; AL, additional terms in the likelihood criterion; CoS, collateral statistics; and WL, weighted likelihood. BH involves SC, stopping according to a criterion, and CO, using a cut‐off level over a completed tree. Item PA covers PP, partition post‐processing, and IC, preprocessing by initialization of centroids. The latter involves DC, distant centroids, and AP, anomalous patterns. PP involves V, variance‐based approach; S, structure‐based approach; C, combining clusterings approach; and R, resampling‐based approach. The latter is composed of SS, sub‐sampling; SD, splitting the data; BS, bootstrapping; and AN, adding noise.

[ Normal View | Magnified View ]

Browse by Topic

Algorithmic Development > Structure Discovery
Technologies > Structure Discovery and Clustering

Access to this WIREs title is by subscription only.

Recommend to Your
Librarian Now!

The latest WIREs articles in your inbox

Sign Up for Article Alerts