Customer and Business Analytics

Chapter 11

K-Centroids Partitioning Cluster Analysis

The last chapter looked at hierarchical clustering, the first form of clustering analysis developed, which continues to be used in many applications. However, hierarchical clustering methods do not scale well to large databases since they require the calculation of the distance between every record in a database, something that works up to a few thousand records, but becomes problematic beyond that point. Partially to deal with this limitation of hierarchical clustering, as well as for a number of other reasons, researchers in several disciplines have developed other methods of cluster analysis since the pioneering work of MacQueen (1967). The clustering approach most commonly used in applied ...