NNPROTOC

Nearest neighbor classifier based on cluster prototypes

   W = NNPROTOC(LABC,A)
   W = LABC*NNPROTOC(A)
   LABOUT = B*W*LABELD

Input
LABC	Index vector, indices of cluster prototypes for M objects.
A	Feature based dataset or double array with M objects (rows)  used for the clustering obtaining LABC.
B	Test objects for which clusters should be found

Output
W	Trained classifier based on the 1NN rule using the prototypes.
LABOUT	Pointers of the objects in B to the cluster prototypes of A.

Description

By this routine new objects B may assigned to the clusters found for  another dataset, A. It uses the 1-nearest neighbor (NN) rule based on the  prototypes found by some clustering applied to A. This is fine for  spherical procedures like kmeans (CLUSTK). The prototypes generated by  procedures like exemplar (CLUSTE), modeseeking (CLUSTM) or mean shift  (CLUSTS) do not represent the cluster shape will. High resolution  clusterings are advised for such procedures.

There is a specific classifier for hierarchical clustering, CLUSTHC. wich  uses the linkage type on which the clusters are based.

These routines generate cluster labels. In case (some) true class labels  are given semi-supervised classifiers can be used, see SEMISUPC and  CLUSTC.

See also

datasets, mappings, knnc, cluste, clusth, clustk, clustm, clusts, clusthc,