Pragmatic AI: An Introduction to Cloud-Based Machine Learning, First Edition

B

Deciding on Cluster Size

There are many examples of k-means clustering in this book. One of the most commonly asked questions about these is how many clusters should be made. There is no correct answer because clustering is the process of creating labels, and two domain experts may use their judgments differently.

In Figure B.1, I created a cluster of the 2013−2014 NBA season stats, and I labeled the eight clusters by descriptions I thought were useful. Another NBA expert may have created fewer or more clusters.

The screen shows 8 graphs from top to bottom that represents Block Role, Limited, Role, Very Good, Role, Elite Handle, Elite Big, ...