kMeansClusteringOnDist

K-means clustering on the distance matrix only

Contents

Syntax

Description

[centerIndex, U, objFun] = kMeansClusteringOnDist(distMat, clusterNum) performs Forgy's k-means clustering on a given distance matrix.

• distMat: distance matrix whose elements are the pairwise distances of the data objects to be clustered
• clusterNum: number of clusters
• centerIndex: final indices of cluster centers
• U: final partition matrix
• objFun: values of the objective function during iterations
• option is an optional argument to control clustering parameters, stopping criteria, and message display during iteration:
  • option.maxIter:Max. number of iteration (default: 100)
  • option.minImprovement:Min. amount of improvement (default 1e-5)
  • option.messageDisplay:Message display during iteration (default: 1)

The clustering process stops when the max. number of iteration is reached, or when the objective function improvement between two consecutive iteration is less than the min. amount of improvement specified.

This function is different from kMeansClustering in that the given input is a distance matrix instead of a data matrix.

Example:

data=dcData(6);

data=data.input;

distMat=distPairwise(data, data);

clusterNum=4;

[centerIndex, U, objFun] = kMeansClusteringOnDist(distMat, clusterNum);

plot(data(1,:), data(2,:), 'ok'); axis image;

maxU = max(U);

for i=1:clusterNum

• index=find(U(i, :) == maxU);
• line(data(1,index), data(2,index), 'linestyle', 'none', 'marker', '*', 'color', getColor(i));

end

Example

data=dcData(6);
data=data.input;
distMat=distPairwise(data, data);
clusterNum=4;
[centerIndex, U, objFun] = kMeansClusteringOnDist(distMat, clusterNum);
plot(data(1,:), data(2,:), 'ok'); axis image;
maxU = max(U);
for i=1:clusterNum
	index=find(U(i, :) == maxU);
	line(data(1,index), data(2,index), 'linestyle', 'none', 'marker', '*', 'color', getColor(i));
end

Iteration count = 1, obj. function = 29.674100
Iteration count = 2, obj. function = 19.388176
Iteration count = 3, obj. function = 12.259786
Iteration count = 4, obj. function = 11.412353
Iteration count = 5, obj. function = 11.337714
Iteration count = 6, obj. function = 11.337714



See Also

kMeansClustering, vqCenterObjInit.


Published with MATLAB® R2020b