Clustering algorithms often want the number of expected cluster prespecified. This can be difficult in some cases, so that detecting the number of clusters during their creation could be sometimes useful. I find that the matter fits into the gravitational theory almost naturally due to the mindset that underpins both field of applications. In the attachment I explain how this can be done. In addition, I exemplify with a comparison with the SAS FASTCLUS procedure on the classic Fisher IRIS data.
------------------------------
[Ulderico] [Santarelli]
[Las Vegas][Nevada]
------------------------------