Open PhDyellow opened 8 months ago
kl and rubin tend to cluster around 20-40 across most datasets when 50 is the upper limit.
kl
rubin
Other metrics tend towards 2-7 or 48-50.
However, not clear if kl and rubin retain their performance when >> 50 clusters are tested.
Other metrics to watch are cindex, friedman, sdindex and sometimes ccc.
cindex
friedman
sdindex
ccc
kl
andrubin
tend to cluster around 20-40 across most datasets when 50 is the upper limit.Other metrics tend towards 2-7 or 48-50.
However, not clear if
kl
andrubin
retain their performance when >> 50 clusters are tested.