Cluster Analysis and Unsupervised Classification

View at publisher Summarise with VIRAT
Authors: David J. Hand
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

The aim of unsupervised classification is to divide a sample or population into classes of similar objects. Cluster analysis is a collection of algorithms which do this. It is useful to distinguish between two different objectives: (i) Imposition or construction, in which the aim is to define a class structure which is convenient or helpful for some purpose, (ii) Discovery or realism, in which the aim is to find natural classes in the data. In both cases, the variables used to describe the objects will determine the possible cluster structures. Notions of similarity, between objects and between clusters, and of the quality of a proposed clustering lie at the heart of the methods. Broadly speaking, there are two types of clustering algorithms. Optimisation algorithms look through the set of candidate clusterings to find one which optimises some score of quality of the cluster structure. Hierarchical methods recursively partition or merge clusters. The k-means algorithm and the single link algorithm are examples of the former and the latter, respectively. Different methods implicitly define what is meant by “cluster” differently, so it is important to identify a method which matches the aims.

Fields of Study & Subjects

Advanced Clustering Algorithms Research Data Mining and Machine Learning Applications Bayesian Methods and Mixture Models

Publication details

Journal / source
Chapman & Hall/CRC eBooks
Pages
186–196
Publication date
27 March 2026
Language
EN
Document type
Book chapter