weka.attributeSelection
Class CfsSubsetEval

java.lang.Object
  weka.attributeSelection.ASEvaluation
      weka.attributeSelection.SubsetEvaluator
          weka.attributeSelection.CfsSubsetEval

All Implemented Interfaces:

OptionHandler, java.io.Serializable

public class CfsSubsetEval

extends SubsetEvaluator

implements OptionHandler

CFS attribute subset evaluator. For more information see:

Hall, M. A. (1998). Correlation-based Feature Subset Selection for Machine Learning. Thesis submitted in partial fulfilment of the requirements of the degree of Doctor of Philosophy at the University of Waikato.

Valid options are: -M
Treat missing values as a seperate value.

-L
Include locally predictive attributes.

Version:

$Revision: 1.19 $

Author:

Mark Hall (mhall@cs.waikato.ac.nz)

See Also:

Serialized Form

Field Summary

private double	m_c_Threshold Threshold for admitting locally predictive features
private int	m_classIndex The class index
private float[][]	m_corr_matrix Holds the matrix of attribute correlations
private Discretize	m_disTransform Discretise attributes when class in nominal
private boolean	m_isNumeric Is the class numeric
private boolean	m_locallyPredictive Include locally predicitive attributes
private boolean	m_missingSeperate Treat missing values as seperate values
private int	m_numAttribs Number of attributes in the training data
private int	m_numInstances Number of instances in the training data
private double[]	m_std_devs Standard deviations of attributes (when using pearsons correlation)
private Instances	m_trainInstances The training instances

Constructor Summary

CfsSubsetEval()
Constructor

Method Summary

private void	addLocallyPredictive(java.util.BitSet best_group)
void	buildEvaluator(Instances data) Generates a attribute evaluator.
private float	correlate(int att1, int att2)
double	evaluateSubset(java.util.BitSet subset) evaluates a subset of attributes
boolean	getLocallyPredictive() Return true if including locally predictive attributes
boolean	getMissingSeperate() Return true is missing is treated as a seperate value
java.lang.String[]	getOptions() Gets the current settings of CfsSubsetEval
java.lang.String	globalInfo() Returns a string describing this attribute evaluator
java.util.Enumeration	listOptions() Returns an enumeration describing the available options.
java.lang.String	locallyPredictiveTipText() Returns the tip text for this property
static void	main(java.lang.String[] args) Main method for testing this class.
java.lang.String	missingSeperateTipText() Returns the tip text for this property
private double	nom_nom(int att1, int att2)
private double	num_nom2(int att1, int att2)
private double	num_num(int att1, int att2)
int[]	postProcess(int[] attributeSet) Calls locallyPredictive in order to include locally predictive attributes (if requested).
protected void	resetOptions()
void	setLocallyPredictive(boolean b) Include locally predictive attributes
void	setMissingSeperate(boolean b) Treat missing as a seperate value
void	setOptions(java.lang.String[] options) Parses and sets a given list of options.
private double	symmUncertCorr(int att1, int att2)
java.lang.String	toString() returns a string describing CFS

Methods inherited from class weka.attributeSelection.ASEvaluation

forName, makeCopies

Methods inherited from class java.lang.Object

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

Field Detail

m_trainInstances

private Instances m_trainInstances

The training instances

m_disTransform

private Discretize m_disTransform

Discretise attributes when class in nominal

m_classIndex

private int m_classIndex

The class index

m_isNumeric

private boolean m_isNumeric

Is the class numeric

m_numAttribs

private int m_numAttribs

Number of attributes in the training data

m_numInstances

private int m_numInstances

Number of instances in the training data

m_missingSeperate

private boolean m_missingSeperate

Treat missing values as seperate values

m_locallyPredictive

private boolean m_locallyPredictive

Include locally predicitive attributes

m_corr_matrix

private float[][] m_corr_matrix

Holds the matrix of attribute correlations

m_std_devs

private double[] m_std_devs

Standard deviations of attributes (when using pearsons correlation)

m_c_Threshold

private double m_c_Threshold

Threshold for admitting locally predictive features

Constructor Detail

CfsSubsetEval

public CfsSubsetEval()

Constructor

Method Detail

globalInfo

public java.lang.String globalInfo()

Returns a string describing this attribute evaluator

Returns:

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

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.

setOptions

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

Parses and sets a given list of options.

Valid options are: -M
Treat missing values as a seperate value.

-L
Include locally predictive attributes.

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

locallyPredictiveTipText

public java.lang.String locallyPredictiveTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

setLocallyPredictive

public void setLocallyPredictive(boolean b)

Include locally predictive attributes

Parameters:

b - true or false

getLocallyPredictive

public boolean getLocallyPredictive()

Return true if including locally predictive attributes

Returns:

true if locally predictive attributes are to be used

missingSeperateTipText

public java.lang.String missingSeperateTipText()

Returns the tip text for this property

Returns:

tip text for this property suitable for displaying in the explorer/experimenter gui

setMissingSeperate

public void setMissingSeperate(boolean b)

Treat missing as a seperate value

Parameters:

b - true or false

getMissingSeperate

public boolean getMissingSeperate()

Return true is missing is treated as a seperate value

Returns:

true if missing is to be treated as a seperate value

getOptions

public java.lang.String[] getOptions()

Gets the current settings of CfsSubsetEval

Specified by:

getOptions in interface OptionHandler

Returns:

an array of strings suitable for passing to setOptions()

buildEvaluator

public void buildEvaluator(Instances data)
                    throws java.lang.Exception

Generates a attribute evaluator. Has to initialize all fields of the evaluator that are not being set via options. CFS also discretises attributes (if necessary) and initializes the correlation matrix.

Specified by:

buildEvaluator in class ASEvaluation

Parameters:

data - set of instances serving as training data

Throws:

java.lang.Exception - if the evaluator has not been generated successfully

evaluateSubset

public double evaluateSubset(java.util.BitSet subset)
                      throws java.lang.Exception

evaluates a subset of attributes

Specified by:

evaluateSubset in class SubsetEvaluator

Parameters:

subset - a bitset representing the attribute subset to be evaluated

Returns:

the "merit" of the subset

Throws:

java.lang.Exception - if the subset could not be evaluated

correlate

private float correlate(int att1,
                        int att2)

symmUncertCorr

private double symmUncertCorr(int att1,
                              int att2)

num_num

private double num_num(int att1,
                       int att2)

num_nom2

private double num_nom2(int att1,
                        int att2)

nom_nom

private double nom_nom(int att1,
                       int att2)

toString

public java.lang.String toString()

returns a string describing CFS

Returns:

the description as a string

addLocallyPredictive

private void addLocallyPredictive(java.util.BitSet best_group)

postProcess

public int[] postProcess(int[] attributeSet)
                  throws java.lang.Exception

Calls locallyPredictive in order to include locally predictive attributes (if requested).

Overrides:

postProcess in class ASEvaluation

Parameters:

attributeSet - the set of attributes found by the search

Returns:

a possibly ranked list of postprocessed attributes

Throws:

java.lang.Exception - if postprocessing fails for some reason

resetOptions

protected void resetOptions()

main

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

Main method for testing this class.

Parameters:

args - the options