Machine Learning Algorithms - Second Edition

Hierarchical Clustering

In this chapter, we're going to discuss a particular clustering technique called Hierarchical Clustering. Instead of working with relationships that exist in the whole dataset, this approach starts with a single entity containing all the elements (divisive) or N separate elements (agglomerative), and proceeds by splitting or merging the clusters according to some specific criteria, which we're going to analyze and compare.

In particular, we are going to discuss the following:

• Hierarchical strategies
• Agglomerative Clustering (metrics, linkages, and dendrograms)
• Connectivity constraints for Agglomerative Clustering