Data Mining Algorithms: Explained Using R

Chapter 12

k-Centers clustering

12.1 Introduction

The clustering task was presented in Section 1.5 as the combination of cluster formation, which identifies similarity-based groups in the training set, and cluster modeling, which creates a model for cluster membership prediction. Dissimilarity-based clustering algorithms address both of these subtasks using measures of instance dissimilarity or similarity. The family of -centers clustering algorithms represents not only the conceptually simplest but also the most popular approach to dissimilarity-based clustering. Of all algorithms using explicit similarity or dissimilarity measures, -centers algorithms employ these measures in the most direct and straightforward way to determine cluster membership.

12.1.1 Basic principle

Algorithms from the -centers family share the same basic operation principle that can be outlined as follows:

1. 1. the number of clusters is predetermined and referred to as (hence the “-” in algorithm names),