edu.ucla.sspace.clustering

Class DirectClustering

java.lang.Object

edu.ucla.sspace.clustering.DirectClustering

All Implemented Interfaces:

Clustering

public class DirectClustering
extends Object
implements Clustering

An implementation of the Direct K-Means clustering available in the CLUTO Clustering Software. This implementation of K-Means varies from the standard implementation in a variety of ways:

After creating the initial set of clusters, each data point is assigned to a cluster if such a move improvies the overall clustering score. This implementation is an iterative method, in that it makes the cluster change immediately and recomputes the centroid after every change.

The clustering objective can be changed by using a different CriterionFunction. The I1Function implements the standard K-Means criterion function. Others try to optimize a variety of different objectives.

Author:

Keith Stevens