clustering

Did you try  pam()  from package "cluster"?

I'd recommend it as an improvement compared to kmeans().

Martin Maechler, ETH Zurich