What methods you use for choosing attributes for kmeans and other clustering (unsupervised) methods?

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view
|

What methods you use for choosing attributes for kmeans and other clustering (unsupervised) methods?

SpaceSeller
What characteristics makes attribute great for kmeans clustering and how do you measure these characteristics? Do you look at the distribution shape, correlation matrix between attributes, pairs graphs, statistical tests...?