In this chapter, we presented Hierarchical Clustering, focusing our attention on the Agglomerative version, which is the only one supported by scikit-learn. We discussed its philosophy, which is rather different to the one adopted by many other methods. In Agglomerative Clustering, the process begins by considering each sample as a single cluster and proceeds by merging the blocks until the number of desired clusters is reached. To perform this task, two elements are needed: a metric function (also called affinity) and a linkage criterion. The former is used to determine the distance between the elements, while the latter is a target function that is used to determine which clusters must be merged.
We also saw how to visualize this process through dendrograms, using SciPy. This technique is quite useful when it's necessary to maintain complete control of the process...
