k-nearest neighbor a supervised data mining method?

Hi. I am doing some research of the k-nearest neighbor method of data mining and it certainly looks like it would fall under the category of a supervised method since pre-made classifications seem to need to exist before starting. Does anyone know of a variation where there are no pre-defined instance groups?

Thanks,
Rf. Received on Tue Mar 21 2006 - 22:59:15 CST