edu.indiana.util.clustering
Class KMedoidClustering

java.lang.Object
  edu.indiana.util.clustering.AbstractClusterMethod
      edu.indiana.util.clustering.KMedoidClustering

All Implemented Interfaces:

ClusterMethod

public class KMedoidClustering
extends AbstractClusterMethod

Performs k-medoid clustering, as follows:
TD(class) = sum(dist(p, case), all cases)
TD = sum(TD(class), all classes)
choose k prototypes
assign remaining cases to nearest prototype
Label A
For each prototype p {
Find the non-prot n for which TD is smallest when n is a protoype instead of p

if the resulting TD is smaller than the TD without swapping {
make n a prot instead of p
reassign the other non-prots
}
}
If you made it all the way through the loop without changing any prototypes, you're done
otherwise goto A

Nested Class Summary

protected class

KMedoidClustering.ECMCopyList

Constructor Summary

KMedoidClustering()

Method Summary

Partition

cluster(Partition partition)

Methods inherited from class edu.indiana.util.clustering.AbstractClusterMethod

cluster, cluster

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

KMedoidClustering

public KMedoidClustering()

Method Detail

cluster

public Partition cluster(Partition partition)