[R] unsupervised random forest classification

[helen.mills at yale.edu]

Friends,
I would like to use Random Forest in unsupervised mode to classify data that are
from a mixture of categorical and continuous variables. The examples I find
online use only continuous data, followed by an mds plot. Is it correct to use
RF to obtain proximities for my mixed dataset and then perform some other form
of clustering (i.e. pam, clara, etc.) using the proximities matrix? Or is there
a way to perform unsupervised classification that will kick out cluster
membership on this dataset in RF itself?

Thanks,
Helen Poulos
Yale School of Forestry