MATLAB: Sort items to existing cluster groups

Question

I applied Y = pdist(X); Z = linkage(Y,'ward'); to cluster my items X. Afterwards, I defined the optimum number of groups. Now, I have additional items Y which I want to sort to the already existing groups received from clustering X. What's the best way to do this? How can I define the 'borders' of the existing groups?

Answer 1

Use the function cluster to give a cluster idx to each sample in X:

 C = cluster(Z,'MaxClust',number_of_groups);

Then find the closest sample for each new observation in W,

C(knnsearch(X,W,'k',1))

You may want also to look at knnclassify for more elaborated voting schemes. See also clusterdata that wraps pdist, linkage and cluster in one function call.