Bisecting K-Means
This is a variant of K-Means
it's a Hierarchical Clustering method, and it's useful for Document Clustering
Algorithm:
start with a single cluster
repeat until have desired number of clusters
choose a cluster to split (e.g. the largest one)
find two subclusters using K-means with k=2k=2 and split
may repeat this procedure several times and take the clusters with highest overall similarity