The Practice of Cluster Analysis

Jon R. Kettenring

Cluster analysis is one of the main methodologies for analyzing multivariate data. While not as neat and tidy as other methods, such as principal components analysis or discriminant analysis, its use is widespread and growing rapidly. The goal of the talk is to document this growth, characterize current usage, provide examples of recent applications, and highlight both good and risky practices. The presentation is based on evidence extracted from database searches and literature reviews covering the years 1995-2004. It will conclude with suggestions for how the current practice of cluster analysis could be improved.