Cluster analysis is a very practical subject. Some 30 years ago, biologists and social scientists began to look for systematic ways to find groups in their data. Because computers were becoming available, the resulting algorithms could actually be implemented. Nowadays clustering methods are applied in many domains, including artificial intelligence and pattern recognition, chemometrics, ecology, economics, the geosciences, marketing, medical research, political science, psychometrics, and many more. This has led to a lot of different methods, and articles on clustering have appeared not only in statistical journals but also in periodicals of all these domains. Clustering is known under a variety of names, such as numerical taxonomy and automatic data classification. Our purpose was to write an applied book for the general user. We wanted to make cluster analysis available to people who do not necessarily have a strong mathematical or statistical background. Rather than giving an extensive survey of clustering methods, leaving the user with a bewildering multitude of methods to choose from, we preferred to select a few methods that together can deal with most applications. This selection was based on a combination of methodological aims (mainly robustness, consistency, and general applicability) and our own experience in applying clustering to a variety of disciplines. The book grew out of several courses on cluster analysis that we taught in Brussels, Delft, and Fribourg. It was extensively tested as a textbook with students of mathematics, biology, economics, and political science. It is one of the few books on cluster analysis containing exercises. The first chapter introduces the main approaches to clustering and provides guidance to the choice between the available methods. It also discusses various types of data (including interval-scaled and binary variables, as well as similarity data) and explains how these can be transformed prior to the actual clustering. The other six chapters each deal with a specific clustering method. These chapters all have the same structure. The first sections give a short description of the clustering method, explain how to use it, and discuss a set of examples. These are followed by two sections (marked with * because they may be skipped without loss of understanding) on the algorithm and its implementation, and on some related methods in the literature. The chapters are relatively independent (except for Chapter 3 which builds on Chapter 2), allowing instructors to cover only one chapter in a statistics course. Another advantage is that researchers can pick out the method they need for their current application, without having to read other chapters. (To achieve this structure, some things that had to be repeated in the text.) Occasionally, we handed a single chapter to someone working on a particular...
