I am working on a problem of unsupervised classification. The data samples are fairly random. Unfortunately, I don't have the luxury of specifying the number of clusters. I would like to use some criteria to automatically identify the number of clusters.

For example, assume there are distinctly 2 clusters in the data points and I use 3 clusters. Is there method to infer that 2 is the right number of clusters vs 3?



Browse other questions tagged or ask your own question.