I don't fully understand why single link techinque in hierachial clustering is not good at handling (discovering) clusters of round or elliptical shapes. Simple examples will help me a lot. Thanks.

  • $\begingroup$ Each different hierarchical linkage method has its own inclinations wrt the shape of a cluster ("cluster metaphor", see pt 3 here). Single linkage method is prone to "chain" and form clusters of irregular, often thread-like curved shapes. The reason for that is obvious. With this method, at any step, two clusters are merged if their closest edges are close enough. No proximity between other parts of the two clusters is taken into consideration. $\endgroup$ – ttnphns May 31 '16 at 12:31
  • 1
    $\begingroup$ (cont.) Therefore clusters that are far apart by their opposite poles easily get merged as being "close". That's why single link method is also called nearest neighbour method. $\endgroup$ – ttnphns May 31 '16 at 12:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.