Pattern Recognition, 4th Edition

Chapter 16. Cluster Validity

16.1. Introduction

A common characteristic of the majority of the clustering algorithms, discussed in the previous chapters, is that they impose a clustering structure on the data set X, even though X may not possess such a structure. In the latter case, the results produced after the application of a clustering algorithm on X are not indicative of the structure of X. In other words, cluster analysis is not a panacea. That is, we must have an indication that the vectors of X form clusters before we apply a clustering algorithm. The problem of verifying whether X possesses a clustering structure, without identifying it explicitly, is known as clustering tendency and is discussed at the end of the chapter.

Let us ...