Elementary Cluster Analysis

The availability of packaged clustering programs means that anyone with data can easily do cluster analysis on it. But many users of this technology dont fully appreciate its many hidden dangers. In todays world of grab and go algorithms, part of my motivation for writing this book is to provide users with a set of cautionary tales about cluster analysis, for it is very much an art as well as a science, and it is easy to stumble if you dont understand its pitfalls. Indeed, it is easy to trip over them even if you do! The parenthetical word usually in the title is very important, because all clustering algorithms can and do fail from time to time.
Modern cluster analysis has become so technically intricate that it is often hard for the beginner or the non-specialist to appreciate and understand its many hidden dangers. Heres how Yogi Berra put it, and he was right: In theory theres no difference between theory and practice. In practice, there is. —Yogi BerraThis book is a step backwards, to four classical methods for clustering in small, static data sets that have all withstood the tests of time.
The overall picture in clustering is quite overwhelming, so any attempt to swim at the deep end of the pool in even a very specialized subfield requires a lot of training. But we all start out at the shallow end (or at least thats where we should start!), and this book is aimed squarely at teaching toddlers not to be afraid of the water. There is no section of this book that, if explored in real depth, cannot be expanded into its own volume. So, if your needs are for an in-depth treatment of all the latest developments in any topic in this volume, the best I can do — what I will try to do anyway — is lead you to the pool, and show you where to jump in.