Unsupervised methods
Johann Bacher, Andreas Pöge, Knut Wenzig · 2021
The basic aim of all clustering methods is to assign objects to groups (clusters) according to similarities in their specific characteristics. The development of clustering methods has varied in intensity and innovation since the 1960s when they first became popular. Today, in the early 21st century, with huge advancements in computer capacity and capability, elaborate and computationally intensive methods have become the norm and the literature has exploded. Clustering methods are available in most statistical software packages, as well as in machine-learning software and data-mining packages. This chapter provides an overview of clustering methods and covers the following topics: (1) steps toward an appropriate cluster solution, (2) clustering methods, (3) criteria to determine the number of clusters, (4) methods to validate cluster solutions, (5) computer programs, (6) application, and (7) summary and recommendations.