HCLUST

hierarchical clustering

   [LABELS, DENDROGRAM] = HCLUST(D,TYPE,K)
    DENDROGRAM = HCLUST(D,TYPE)

Input
D	dissimilarity matrix
TYPE	string name of clustering criterion (optional) -  's' or 'single' : single linkage (default) -  'c' or 'complete' : complete linkage -  'a' or 'average' : average linkage (weighted over cluster sizes)
K	number of clusters (optional)

Output
LABELS	vector with labels
DENDROGRAM	matrix with dendrogram

Description

Computation of cluster labels and a dendrogram between the clusters for  objects with a given distance matrix D. K is the desired number of  clusters. The dendrogram is a 2*K matrix. The first row yields all  cluster sizes. The second row is the cluster level on which the set of  clusters starting at that position is merged with the set of clusters  just above it in the dendrogram. A dendrogram may be plotted by PLOTDG.

    DENDROGRAM = HCLUST(D,TYPE)

As in this case no clustering level is supplied, just the entire  dendrogram is returned. The first row now contains the object indices.

Example(s)

 a = gendats([25 25],20,5);    % 50 points in 20 dimensional feature space
 d = sqrt(distm(a));           % Euclidean distances
 dendg = hclust(d,'complete'); % dendrogram
 plotdg(dendg)
 lab = hclust(d,'complete',2); % labels
 confmat(lab,getlabels(a));    % confusion matrix

See also

plotdg, kmeans, kcentres, modeseek, emclust,