Cluster Summaries in CLUTO
I have been working on clustering problems for a few months now. I just started playing with CLUTO and find it quite neat. However, I am wondering about the numbers that are generated by the -showfeatures option.
The manual says that a "descriptive feature" is measured as the percentage of average similarity between objects of the cluster. Can somebody please provide a formula for this? Also for the discriminating features? I'd greatly appreciate it!