One issue is how to measure the quality of the results provided by k-means classification. We use some internal measures:
A good classification has:
Low within class variance -- compactness
High distance between class centers -- isolation
This page is maintained
by Prof. Miriam Leeser