de.aitools.dm.clustering.algorithms
Class DBScan

java.lang.Object
  de.aitools.dm.clustering.algorithms.ASoftClusterer
      de.aitools.dm.clustering.algorithms.AClusterer
          de.aitools.dm.clustering.algorithms.DBScan

All Implemented Interfaces:

Clusterer, SoftClusterer

public final class DBScan
extends AClusterer

This class is experimental at best. As you can see, it uses deprecated classes. It is also not tested well and I do not provide support for it. It has passes some basic tests (see the test folder). It also seemed to work on some real dataset.

The DBScanConfiguration is, however, not finished. The algorithm has two parameters, the number of neighborhood points needed to label a point as a Core Point, and the range which defines what the neighborhood is.

In short: The algorithm labels all points with at least coreMinNeighbors other points within neighborhoodProximity as core points. All other points with a core point in their neighborhood are called border points. The remaining points are called noise. Clusters are formed by connecting core points with all other core points in their neighborhood and the assigning the border points to the nearest cluster.

The cluster label 0 is reserved for the noise points.

There are some ways to calculate one or both parameters from the data (for which the DBScanConfiguration interface was created). This is, however, not finished. Not even partly.

You can use this implementation to try things out. But if you plan to use it seriously, I would recommend to reimplement it.

Version:

$Id: DBScan.java,v 1.2 2011/11/09 13:40:46 hoppe Exp $

Author:

johannes.kiesel(/\t)uni-weimar.de

Nested Class Summary

class	DBScan.DBScanAllParameters TODO
static interface	DBScan.DBScanConfiguration TODO
class	DBScan.DBScanGraphAverageSlopeOutlierDetector TODO
static interface	DBScan.DBScanGraphKneeDetector TODO
class	DBScan.DBScanSearchProximity TODO

Field Summary

Fields inherited from interface de.aitools.dm.clustering.Clusterer

DEFAULT_SEED

Constructor Summary

DBScan(Configuration configuration)
TODO

DBScan(Proximity<Vector> proximityMeasure)
TODO

Method Summary

int[]	cluster(Vector[] data)
int[]	cluster(Vector[] data, int coreMinNeighbors) TODO
int[]	cluster(Vector[] data, int coreMinNeighbors, DBScan.DBScanGraphKneeDetector neighborhoodProximityDetector) TODO
int[]	cluster(Vector[] data, int coreMinNeighbors, double neighborhoodProximity) TODO
static void	main(java.lang.String[] args)
void	setClusterConfiguration(DBScan.DBScanConfiguration configuration) TODO
void	setProximityMeasure(Proximity<Vector> proximityMeasure) Sets the proximity measure to be used for clustering.

Methods inherited from class de.aitools.dm.clustering.algorithms.AClusterer

cluster, cluster, cluster, clusterSoft

Methods inherited from class de.aitools.dm.clustering.algorithms.ASoftClusterer

clusterSoft, clusterSoft, clusterSoft, getBiggestRange

Methods inherited from class java.lang.Object

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

Constructor Detail

DBScan

public DBScan(Proximity<Vector> proximityMeasure)

TODO

Parameters:

proximityMeasure -

DBScan

public DBScan(Configuration configuration)

TODO

Parameters:

configuration -

Method Detail

setProximityMeasure

public void setProximityMeasure(Proximity<Vector> proximityMeasure)

Sets the proximity measure to be used for clustering.

Parameters:

proximityMeasure - The measure to use.

setClusterConfiguration

public void setClusterConfiguration(DBScan.DBScanConfiguration configuration)

TODO

Parameters:

configuration -

cluster

public int[] cluster(Vector[] data)

Specified by:

cluster in interface Clusterer

Specified by:

cluster in class AClusterer

cluster

public int[] cluster(Vector[] data,
                     int coreMinNeighbors,
                     double neighborhoodProximity)

TODO

Parameters:

data -

coreMinNeighbors -

neighborhoodProximity -

Returns:

cluster

public int[] cluster(Vector[] data,
                     int coreMinNeighbors)

TODO

Parameters:

data -

coreMinNeighbors -

Returns:

cluster

public int[] cluster(Vector[] data,
                     int coreMinNeighbors,
                     DBScan.DBScanGraphKneeDetector neighborhoodProximityDetector)

TODO

Parameters:

data -

coreMinNeighbors -

neighborhoodProximityDetector -

Returns:

main

public static void main(java.lang.String[] args)

Parameters:

args -