How does K-means work in binary variable?

johnny5550822

Hi,

I know traditional k-means (distance one) does not work well in binary variable. I installed the weka extension in rapidminer and use the k-means (kernel one); my dataset (medical ) contains all binary variables. Surprisingly, it work well such that when k=2 with dot-kernel, it separate 2 groups of clasuter which one group has almost none of the disease/medication, and another group has at least 2 disease/medication.

Should I use k-means to cluster data with only binary feature in rapidminer?

Thanks!