0

I'm looking for a method(s) to compare the fit of k-Medoids and a GMM.

Currently, I'm looking at the distance between the max-min means of the GMM clusters and comparing that to the max-min medoid locations of the k-Medoids.

Is this a good method or is there better, more accurate methods?

It's a timeseries dataset.

enter image description here

For example, 'max' is the center of the blue cluster and 'min' in the center of the red cluster.

Andrew Feenan
  • 1
  • 1
  • Please clarify your specific problem or provide additional details to highlight exactly what you need. As it's currently written, it's hard to tell exactly what you're asking. – Lucas Morin Jun 25 '23 at 11:41

1 Answers1

0

I presume 'max' means the maximum distance of an observation from its cluster center? And then 'min' would be by definition 0 for K-Medoids?

IF that is the case it would seem to be excluding much of the information that a basic RMSE would give you. By RMSE here I mean the root of the sum of the squares of the distances between the observations and the "center" of their assigned cluster. The RMSE (or MAD if the situation warrants it) does a pretty good job so to choose a different metric clear evidence should be provided

WestCoastProjects
  • 191
  • 9
  • I have edited my question so show what I mean. By 'max' I mean the top cluster and by 'min I mean the bottom cluster, I would find the difference between then for a max-min value, then compare the max-min for GMM and k-medoids. – Andrew Feenan Jun 26 '23 at 10:48