Chapter 7 Advanced Clustering Methods

This third chapter in the treatment of classic clustering methods considers two more advanced partitioning clustering techniques. The first, K-Medians, is a straightforward variant on K-Means. It operates in exactly the same manner, except that the representative center of each cluster is the median instead of the mean.

The second method is similar to K-Medians and is often confused with it. K-Medoids is a method where the center of each cluster is an actual cluster member, in contrast to what holds for both K-Means and K-Medians.

Both methods operate in largely the same manner as K-Means, but they differ in the way the central point of each cluster is defined and the manner in which the nearest points are assigned.

Each method will be covered in turn. However since their implementation shares a lot of commonalities with the methods covered in Chapters 5 and 6, the focus will be only on those aspects that differ.

The Chicago SDOH sample data set is again used to illustrate the techniques.