moa.clusterers

Class KMeans

java.lang.Object

moa.clusterers.KMeans

public class KMeans
extends Object

A kMeans implementation for microclusterings. For now it only uses the real centers of the groundtruthclustering for implementation. There should also be an option to use random centers. TODO: random centers TODO: Create a macro clustering interface to make different macro clustering algorithms available to micro clustering algorithms like clustream, denstream and clustree

Constructor Summary

Constructors

Constructor and Description
KMeans()

Method Summary

Methods

Modifier and Type	Method and Description
static Clustering	gaussianMeans(Clustering gtClustering, Clustering clustering)
static Clustering	kMeans(Cluster[] centers, List<? extends Cluster> data) This kMeans implementation clusters a big number of microclusters into a smaller amount of macro clusters.

Methods inherited from class java.lang.Object

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

Constructor Detail

KMeans

public KMeans()

Method Detail

kMeans

public static Clustering kMeans(Cluster[] centers,
                List<? extends Cluster> data)

This kMeans implementation clusters a big number of microclusters into a smaller amount of macro clusters. To make it comparable to other algorithms it uses the real centers of the ground truth macro clustering to have the best possible initialization. The quality of resulting macro clustering yields an upper bound for kMeans on the underlying microclustering.

Parameters:

centers - of the ground truth clustering

data - list of microclusters

Returns:

gaussianMeans

public static Clustering gaussianMeans(Clustering gtClustering,
                       Clustering clustering)