jsat.clustering

Class HDBSCAN

java.lang.Object

jsat.clustering.HDBSCAN

All Implemented Interfaces:

java.io.Serializable, Clusterer, Parameterized

public class HDBSCAN
extends java.lang.Object
implements Clusterer, Parameterized

HDBSCAN is a density based clustering algorithm that is an improvement over DBSCAN. Unlike its predecessor, HDBSCAN works with variable density datasets and does not need a search radius to be specified. The original paper presents HDBSCAN with two parameters mpts and mclSize, but recomends that they can be set to the same value and effectively behave as if only one parameter exists. This implementation allows for setting both independtly, but the single parameter constructors will use the same value for both parameters.
NOTE: The current implementation has O(N²) run time, though this may be improved in the future with more advanced algorithms.

See: Campello, R. J. G. B., Moulavi, D., & Sander, J. (2013). Density-Based Clustering Based on Hierarchical Density Estimates. In J. Pei, V. Tseng, L. Cao, H. Motoda, & G. Xu (Eds.), Advances in Knowledge Discovery and Data Mining (pp. 160–172). Springer Berlin Heidelberg. doi:10.1007/978-3-642-37456-2_14

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
HDBSCAN() Creates a new HDBSCAN object using a threshold of 15 points to form a cluster.
HDBSCAN(DistanceMetric dm, int m_pts) Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.
HDBSCAN(DistanceMetric dm, int m_pts, int m_clSize, VectorCollection<Vec> vc) Creates a new HDBSCAN using the full specification of the algorithm, where two parameters may be altered.
HDBSCAN(DistanceMetric dm, int m_pts, VectorCollection<Vec> vcf) Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.
HDBSCAN(HDBSCAN toCopy) Copy constructor
HDBSCAN(int m_pts) Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.

Method Summary

Modifier and Type	Method and Description
HDBSCAN	clone()
int[]	cluster(DataSet dataSet, boolean parallel, int[] designations) Performs clustering on the given data set.
DistanceMetric	getDistanceMetrics()
int	getMinClusterSize()
int	getMinPoints()
void	setDistanceMetrics(DistanceMetric dm) Sets the distance metric to use for determining closeness between data points
void	setMinClusterSize(int m_clSize)
void	setMinPoints(int m_pts)

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.clustering.Clusterer

cluster, cluster, cluster, supportsWeightedData

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

HDBSCAN

public HDBSCAN()

Creates a new HDBSCAN object using a threshold of 15 points to form a cluster.

HDBSCAN

public HDBSCAN(int m_pts)

Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.

Parameters:

m_pts - the minimum number of points needed to form a cluster and the number of neighbors to consider

HDBSCAN

public HDBSCAN(DistanceMetric dm,
               int m_pts)

Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.

Parameters:

dm - the distance metric to use for finding nearest neighbors

m_pts - the minimum number of points needed to form a cluster and the number of neighbors to consider

HDBSCAN

public HDBSCAN(DistanceMetric dm,
               int m_pts,
               VectorCollection<Vec> vcf)

Creates a new HDBSCAN using the simplified form, where the only parameter is a single value.

Parameters:

dm - the distance metric to use for finding nearest neighbors

m_pts - the minimum number of points needed to form a cluster and the number of neighbors to consider

vcf - the vector collection to use for accelerating nearest neighbor queries

HDBSCAN

public HDBSCAN(DistanceMetric dm,
               int m_pts,
               int m_clSize,
               VectorCollection<Vec> vc)

Creates a new HDBSCAN using the full specification of the algorithm, where two parameters may be altered. In the simplified version both parameters always have the same value.

Parameters:

dm - the distance metric to use for finding nearest neighbors

m_pts - the number of neighbors to consider, acts as a smoothing over the density estimate

m_clSize - the minimum number of data points needed to form a cluster

vc - the vector collection to use for accelerating nearest neighbor queries

HDBSCAN

public HDBSCAN(HDBSCAN toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

setMinClusterSize

public void setMinClusterSize(int m_clSize)

Parameters:

m_clSize - the minimum number of data points needed to form a cluster

getMinClusterSize

public int getMinClusterSize()

Returns:

the minimum number of data points needed to form a cluster

setDistanceMetrics

public void setDistanceMetrics(DistanceMetric dm)

Sets the distance metric to use for determining closeness between data points

Parameters:

dm - the distance metric to determine nearest neighbors with

getDistanceMetrics

public DistanceMetric getDistanceMetrics()

Returns:

the distance metric to determine nearest neighbors with

setMinPoints

public void setMinPoints(int m_pts)

Parameters:

m_pts - the number of neighbors to consider, acts as a smoothing over the density estimate

getMinPoints

public int getMinPoints()

Returns:

the number of neighbors to consider, acts as a smoothing over the density estimate

clone

public HDBSCAN clone()

Specified by:

clone in interface Clusterer

Overrides:

clone in class java.lang.Object

cluster

public int[] cluster(DataSet dataSet,
                     boolean parallel,
                     int[] designations)

Description copied from interface: Clusterer

Performs clustering on the given data set. Parameters may be estimated by the method, or other heuristics performed.

Specified by:

cluster in interface Clusterer

Parameters:

dataSet - the data set to perform clustering on

parallel - true if multiple threads should be used to perform clustering. false if it should be done in a single threaded manner.

designations - the array which will contain the designated values. The array will be altered and returned by the function. If null is given, a new array will be created and returned.

Returns:

the int[]