weka.datagenerators
Class BIRCHCluster

java.lang.Object
  weka.datagenerators.ClusterGenerator
      weka.datagenerators.BIRCHCluster

All Implemented Interfaces:

OptionHandler, java.io.Serializable

public class BIRCHCluster

extends ClusterGenerator

implements OptionHandler, java.io.Serializable

Cluster data generator designed for the BIRCH System Dataset is generated with instances in K clusters. Instances are 2-d data points. Each cluster is characterized by the number of data points in it its radius and its center. The location of the cluster centers is determined by the pattern parameter. Three patterns are currently supported grid, sine and random. todo: (out of: BIRCH: An Efficient Data Clustering Method for Very Large Databases; T. Zhang, R. Ramkrishnan, M. Livny; 1996 ACM) Class to generate data randomly by producing a decision list. The decision list consists of rules. Instances are generated randomly one by one. If decision list fails to classify the current instance, a new rule according to this current instance is generated and added to the decision list.

The option -V switches on voting, which means that at the end of the generation all instances are reclassified to the class value that is supported by the most rules.

This data generator can generate 'boolean' attributes (= nominal with the values {true, false}) and numeric attributes. The rules can be 'A' or 'NOT A' for boolean values and 'B < random_value' or 'B >= random_value' for numeric values.

Valid options are:

-G
The pattern for instance generation is grid.
This flag cannot be used at the same time as flag I. The pattern is random, if neither flag G nor flag I is set.

-I
The pattern for instance generation is sine.
This flag cannot be used at the same time as flag G. The pattern is random, if neither flag G nor flag I is set.

-N num .. num
The range of the number of instances in each cluster (default 1..50).
Lower number must be between 0 and 2500, upper number must be between 50 and 2500.

-R num .. num
The range of the radius of the clusters (default 0.1 .. SQRT(2)).
Lower number must be between 0 and SQRT(2), upper number must be between
SQRT(2) and SQRT(32).

-M num
Distance multiplier, only used if pattern is grid (default 4).

-C num
Number of cycles, only used if pattern is sine (default 4).

-O
Flag for input order is ordered. If flag is not set then input order is randomized.

-P num
Noise rate in percent. Can be between 0% and 30% (default 0%).
(Remark: The original algorithm only allows noise up to 10%.)

-S seed
Random number seed for random function used (default 1).

Version:

$Revision: 1.2 $

Author:

Gabi Schmidberger (gabi@cs.waikato.ac.nz)

See Also:

Serialized Form

Nested Class Summary

private class	BIRCHCluster.Cluster class to represent cluster
private class	BIRCHCluster.GridVector class to represent Vector for placement of the center in space

Field Summary

static int	GRID
private FastVector	m_ClusterList
private Instances	m_DatasetFormat
private int	m_Debug
private double	m_DistMult
private int	m_GridSize
private double	m_GridWidth
private int	m_InputOrder
private int	m_MaxInstNum
private double	m_MaxRadius
private int	m_MinInstNum
private double	m_MinRadius
private double	m_NoiseRate
private int	m_NumCycles
private int	m_Pattern
private java.util.Random	m_Random
private int	m_Seed
static int	ORDERED
static int	RANDOM
static int	RANDOMIZED
static int	SINE

Fields inherited from class weka.datagenerators.ClusterGenerator

m_NumAttributes, m_NumClusters

Constructor Summary

BIRCHCluster()

Method Summary

private FastVector	defineClusters(java.util.Random random) Defines the clusters
private FastVector	defineClustersGRID(java.util.Random random) Defines the clusters if pattern is GRID
private FastVector	defineClustersRANDOM(java.util.Random random) Defines the clusters if pattern is RANDOM
Instances	defineDataFormat() Initializes the format for the dataset produced.
Instance	generateExample() Generate an example of the dataset.
Instances	generateExamples() Generate all examples of the dataset.
Instances	generateExamples(java.util.Random random, Instances format) Generate all examples of the dataset.
java.lang.String	generateFinished() Compiles documentation about the data generation after the generation process
private Instance	generateInstance(Instances format, java.util.Random randomG, double stdDev, double[] center, java.lang.String cName) Generate an example of the dataset.
java.lang.String	generateStart() Compiles documentation about the data generation before the generation process
Instances	getDatasetFormat() Gets the dataset format.
double	getDistMult() Gets the distance multiplier.
boolean	getGridFlag() Gets the grid flag (option G).
int	getInputOrder() Gets the input order.
java.lang.String	getInstNums() Gets the upper and lower boundary for instances per cluster.
int	getMaxInstNum() Gets the upper boundary for instances per cluster.
double	getMaxRadius() Gets the upper boundary for the radiuses of the clusters.
int	getMinInstNum() Gets the lower boundary for instances per cluster.
double	getMinRadius() Gets the lower boundary for the radiuses of the clusters.
double	getNoiseRate() Gets the percentage of noise set.
int	getNumCycles() Gets the number of cycles.
java.lang.String[]	getOptions() Gets the current settings of the datagenerator BIRCHCluster.
boolean	getOrderedFlag() Gets the ordered flag (option O).
int	getPattern() Gets the pattern type.
java.lang.String	getRadiuses() Gets the upper and lower boundary for the radius of the clusters.
java.util.Random	getRandom() Gets the random generator.
int	getSeed() Gets the random number seed.
boolean	getSineFlag() Gets the sine flag (option S).
boolean	getSingleModeFlag() Gets the single mode flag.
java.lang.String	globalInfo() Returns a string describing this data generator.
java.util.Enumeration	listOptions() Returns an enumeration describing the available options.
static void	main(java.lang.String[] argv) Main method for testing this class.
void	setDatasetFormat(Instances newDatasetFormat) Sets the dataset format.
void	setDefaultOptions() Sets all options to their default values.
void	setDistMult(double newDistMult) Sets the distance multiplier.
void	setInputOrder(int newInputOrder) Sets the input order.
void	setInstNums(java.lang.String fromTo) Sets the upper and lower boundary for instances per cluster.
void	setMaxInstNum(int newMaxInstNum) Sets the upper boundary for instances per cluster.
void	setMaxRadius(double newMaxRadius) Sets the upper boundary for the radiuses of the clusters.
void	setMinInstNum(int newMinInstNum) Sets the lower boundary for instances per cluster.
void	setMinRadius(double newMinRadius) Sets the lower boundary for the radiuses of the clusters.
void	setNoiseRate(double newNoiseRate) Sets the percentage of noise set.
void	setNumCycles(int newNumCycles) Sets the the number of cycles.
void	setOptions(java.lang.String[] options) Parses a list of options for this object.
void	setPattern(int newPattern) Sets the pattern type.
void	setRadiuses(java.lang.String fromTo) Sets the upper and lower boundary for the radius of the clusters.
void	setRandom(java.util.Random newRandom) Sets the random generator.
void	setSeed(int newSeed) Sets the random number seed.

Methods inherited from class weka.datagenerators.ClusterGenerator

getClassFlag, getDebug, getFormat, getNumAttributes, getNumClusters, getNumExamplesAct, getOutput, getRelationName, makeData, setClassFlag, setDebug, setFormat, setNumAttributes, setNumClusters, setNumExamplesAct, setOutput, setRelationName, toStringFormat

Methods inherited from class java.lang.Object

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

Field Detail

m_MinInstNum

private int m_MinInstNum

m_MaxInstNum

private int m_MaxInstNum

m_MinRadius

private double m_MinRadius

m_MaxRadius

private double m_MaxRadius

GRID

public static final int GRID

See Also:

Constant Field Values

SINE

public static final int SINE

See Also:

Constant Field Values

RANDOM

public static final int RANDOM

See Also:

Constant Field Values

m_Pattern

private int m_Pattern

m_DistMult

private double m_DistMult

m_NumCycles

private int m_NumCycles

ORDERED

public static final int ORDERED

See Also:

Constant Field Values

RANDOMIZED

public static final int RANDOMIZED

See Also:

Constant Field Values

m_InputOrder

private int m_InputOrder

m_NoiseRate

private double m_NoiseRate

m_Seed

private int m_Seed

m_DatasetFormat

private Instances m_DatasetFormat

m_Random

private java.util.Random m_Random

m_Debug

private int m_Debug

m_ClusterList

private FastVector m_ClusterList

m_GridSize

private int m_GridSize

m_GridWidth

private double m_GridWidth

Constructor Detail

BIRCHCluster

public BIRCHCluster()

Method Detail

globalInfo

public java.lang.String globalInfo()

Returns a string describing this data generator.

Returns:

a description of the data generator suitable for displaying in the explorer/experimenter gui

setInstNums

public void setInstNums(java.lang.String fromTo)

Sets the upper and lower boundary for instances per cluster.

getInstNums

public java.lang.String getInstNums()

Gets the upper and lower boundary for instances per cluster.

Returns:

the string containing the upper and lower boundary for instances per cluster separated by ..

getMinInstNum

public int getMinInstNum()

Gets the lower boundary for instances per cluster.

Returns:

the the lower boundary for instances per cluster

setMinInstNum

public void setMinInstNum(int newMinInstNum)

Sets the lower boundary for instances per cluster.

Parameters:

newMinInstNum - new lower boundary for instances per cluster

getMaxInstNum

public int getMaxInstNum()

Gets the upper boundary for instances per cluster.

Returns:

the upper boundary for instances per cluster

setMaxInstNum

public void setMaxInstNum(int newMaxInstNum)

Sets the upper boundary for instances per cluster.

Parameters:

newMaxInstNum - new upper boundary for instances per cluster

setRadiuses

public void setRadiuses(java.lang.String fromTo)

Sets the upper and lower boundary for the radius of the clusters.

getRadiuses

public java.lang.String getRadiuses()

Gets the upper and lower boundary for the radius of the clusters.

Returns:

the string containing the upper and lower boundary for the radius of the clusters, separated by ..

getMinRadius

public double getMinRadius()

Gets the lower boundary for the radiuses of the clusters.

Returns:

the lower boundary for the radiuses of the clusters

setMinRadius

public void setMinRadius(double newMinRadius)

Sets the lower boundary for the radiuses of the clusters.

Parameters:

newMinRadius - new lower boundary for the radiuses of the clusters

getMaxRadius

public double getMaxRadius()

Gets the upper boundary for the radiuses of the clusters.

Returns:

the upper boundary for the radiuses of the clusters

setMaxRadius

public void setMaxRadius(double newMaxRadius)

Sets the upper boundary for the radiuses of the clusters.

Parameters:

newMaxRadius - new upper boundary for the radiuses of the clusters

getGridFlag

public boolean getGridFlag()

Gets the grid flag (option G).

Returns:

true if grid flag is set

getSineFlag

public boolean getSineFlag()

Gets the sine flag (option S).

Returns:

true if sine flag is set

getPattern

public int getPattern()

Gets the pattern type.

Returns:

the current pattern type

setPattern

public void setPattern(int newPattern)

Sets the pattern type.

Parameters:

newPattern - new pattern type

getDistMult

public double getDistMult()

Gets the distance multiplier.

Returns:

the distance multiplier

setDistMult

public void setDistMult(double newDistMult)

Sets the distance multiplier.

Parameters:

newDistMult - new distance multiplier

getNumCycles

public int getNumCycles()

Gets the number of cycles.

Returns:

the number of cycles

setNumCycles

public void setNumCycles(int newNumCycles)

Sets the the number of cycles.

Parameters:

newNumCycles - new number of cycles

getInputOrder

public int getInputOrder()

Gets the input order.

Returns:

the current input order

setInputOrder

public void setInputOrder(int newInputOrder)

Sets the input order.

Parameters:

newInputOrder - new input order

getOrderedFlag

public boolean getOrderedFlag()

Gets the ordered flag (option O).

Returns:

true if ordered flag is set

getNoiseRate

public double getNoiseRate()

Gets the percentage of noise set.

Returns:

the percentage of noise set

setNoiseRate

public void setNoiseRate(double newNoiseRate)

Sets the percentage of noise set.

Parameters:

newNoiseRate - new percentage of noise

getRandom

public java.util.Random getRandom()

Gets the random generator.

Returns:

the random generator

setRandom

public void setRandom(java.util.Random newRandom)

Sets the random generator.

Parameters:

newRandom - is the random generator.

getSeed

public int getSeed()

Gets the random number seed.

Returns:

the random number seed.

setSeed

public void setSeed(int newSeed)

Sets the random number seed.

Parameters:

newSeed - the new random number seed.

getDatasetFormat

public Instances getDatasetFormat()

Gets the dataset format.

Returns:

the dataset format.

setDatasetFormat

public void setDatasetFormat(Instances newDatasetFormat)

Sets the dataset format.

Parameters:

newDatasetFormat - the new dataset format.

getSingleModeFlag

public boolean getSingleModeFlag()

Gets the single mode flag.

Specified by:

getSingleModeFlag in class ClusterGenerator

Returns:

true if methode generateExample can be used.

listOptions

public java.util.Enumeration listOptions()

Returns an enumeration describing the available options.

Specified by:

listOptions in interface OptionHandler

Returns:

an enumeration of all the available options

setDefaultOptions

public void setDefaultOptions()

Sets all options to their default values.

setOptions

public void setOptions(java.lang.String[] options)
                throws java.lang.Exception

Parses a list of options for this object.

For list of valid options see class description.

Specified by:

setOptions in interface OptionHandler

Parameters:

options - the list of options as an array of strings

Throws:

java.lang.Exception - if an option is not supported

getOptions

public java.lang.String[] getOptions()

Gets the current settings of the datagenerator BIRCHCluster.

Specified by:

getOptions in interface OptionHandler

Returns:

an array of strings suitable for passing to setOptions

defineDataFormat

public Instances defineDataFormat()
                           throws java.lang.Exception

Initializes the format for the dataset produced.

Specified by:

defineDataFormat in class ClusterGenerator

Returns:

the output data format

Throws:

java.lang.Exception - data format could not be defined

generateExample

public Instance generateExample()
                         throws java.lang.Exception

Generate an example of the dataset.

Specified by:

generateExample in class ClusterGenerator

Returns:

the instance generated

Throws:

java.lang.Exception - if format not defined or generating
examples one by one is not possible, because voting is chosen

generateExamples

public Instances generateExamples()
                           throws java.lang.Exception

Generate all examples of the dataset.

Specified by:

generateExamples in class ClusterGenerator

Returns:

the instance generated

Throws:

java.lang.Exception - if format not defined

generateExamples

public Instances generateExamples(java.util.Random random,
                                  Instances format)
                           throws java.lang.Exception

Generate all examples of the dataset.

Returns:

the instance generated

Throws:

java.lang.Exception - if format not defined

generateInstance

private Instance generateInstance(Instances format,
                                  java.util.Random randomG,
                                  double stdDev,
                                  double[] center,
                                  java.lang.String cName)

Generate an example of the dataset.

Returns:

the instance generated

Throws:

java.lang.Exception - if format not defined or generating
examples one by one is not possible, because voting is chosen

defineClusters

private FastVector defineClusters(java.util.Random random)
                           throws java.lang.Exception

Defines the clusters

Parameters:

random - random number generator

Throws:

java.lang.Exception

defineClustersGRID

private FastVector defineClustersGRID(java.util.Random random)
                               throws java.lang.Exception

Defines the clusters if pattern is GRID

Parameters:

random - random number generator

Throws:

java.lang.Exception

defineClustersRANDOM

private FastVector defineClustersRANDOM(java.util.Random random)
                                 throws java.lang.Exception

Defines the clusters if pattern is RANDOM

Parameters:

random - random number generator

Throws:

java.lang.Exception

generateFinished

public java.lang.String generateFinished()
                                  throws java.lang.Exception

Compiles documentation about the data generation after the generation process

Specified by:

generateFinished in class ClusterGenerator

Returns:

string with additional information about generated dataset

Throws:

java.lang.Exception - no input structure has been defined

generateStart

public java.lang.String generateStart()

Compiles documentation about the data generation before the generation process

Specified by:

generateStart in class ClusterGenerator

Returns:

string with additional information

main

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

Main method for testing this class.

Parameters:

argv - should contain arguments for the data producer: