0

Given a set of points, I'm trying to find the right cluster. However, I am lost on what the process is. Here is the graph of all possible points.

enter image description here

I am unsure what I should look at

  • "gap statistic" is a good search term to learn about this. – whuber Jan 17 '23 at 18:36
  • 1
    There is no single right number of clusters or right clustering. There is no unique definition of what a cluster is, and depending on the definition, there can be different legitimate clusterings on the same data set. So the user needs to specify what kind of clusters they want. For example, in your data, some would say one-point clusters are fine, and some wouldn't count a single point as a cluster. More background: https://arxiv.org/abs/1503.02059 – Christian Hennig Jan 17 '23 at 21:07
  • 1
    There is a large number of approaches for determining the number of clusters, but they may give different results on the same data. Some search terms: Gap statistic, average silhouette width, prediction strength, Calinski-Harabasz index, Sugar-James distortion, bootstrap stability. – Christian Hennig Jan 17 '23 at 21:10

0 Answers0