scikit-learn : Machine Learning Simplified

Alternative clustering methods

The scikit-learn toolkit includes several clustering algorithms, all of them including similar methods and parameters to those we used in k-means. In this section we will briefly review some of them, suggesting some of their advantages.

A typical problem for clustering is that most methods require the number of clusters we want to identify. The general approach to solve this is to try different numbers and let an expert determine which works best using techniques such as dimensionality reduction to visualize clusters. There are also some methods that try to automatically calculate the number of clusters. Scikit-learn includes an implementation of Affinity Propagation, a method that looks for instances that are the ...