K-means requires one to specify the number of clusters; clustering based on HCL requires the user to visually inspect a tree. Have you successfully used an automatic clustering algorithm that provides an optimal partitioning of clustered data? Thanks, Anjan
Looked here?