weka.attributeSelection
Class PrincipalComponents

java.lang.Object
  weka.attributeSelection.ASEvaluation
      weka.attributeSelection.AttributeEvaluator
          weka.attributeSelection.UnsupervisedAttributeEvaluator
              weka.attributeSelection.PrincipalComponents

All Implemented Interfaces:

AttributeTransformer, OptionHandler, java.io.Serializable

public class PrincipalComponents

extends UnsupervisedAttributeEvaluator

implements AttributeTransformer, OptionHandler

Class for performing principal components analysis/transformation.

Valid options are:

-N
Don't normalize the input data.

-R
Retain enough pcs to account for this proportion of the variance.

-T
Transform through the PC space and back to the original space.

Version:

$Revision: 1.25 $

Author:

Mark Hall (mhall@cs.waikato.ac.nz), Gabi Schmidberger (gabi@cs.waikato.ac.nz)

See Also:

Serialized Form

Field Summary

private Remove	m_attribFilter used to remove the class column if a class column is set
private Remove	m_attributeFilter
private int	m_classIndex Class index
private double[][]	m_correlation Correlation matrix for the original data
private double	m_coverVariance the amount of varaince to cover in the original data when retaining the best n PC's
private double[]	m_eigenvalues Eigenvalues for the corresponding eigenvectors
private double[][]	m_eigenvectors Will hold the unordered linear transformations of the (normalized) original data
private double[][]	m_eTranspose holds the transposed eigenvectors for converting back to the original space
private boolean	m_hasClass Data has a class set
private NominalToBinary	m_nominalToBinFilter
private boolean	m_normalize normalize the input data?
private Normalize	m_normalizeFilter
private int	m_numAttribs Number of attributes
private int	m_numInstances Number of instances
private Instances	m_originalSpaceFormat The header for data transformed back to the original space
private int	m_outputNumAtts The number of attributes in the pc transformed data
private ReplaceMissingValues	m_replaceMissingFilter Filters for original data
private int[]	m_sortedEigens Sorted eigenvalues
private double	m_sumOfEigenValues sum of the eigenvalues
private Instances	m_trainCopy Keep a copy for the class attribute (if set)
private Instances	m_trainInstances The data to transform analyse/transform
private boolean	m_transBackToOriginal transform the data through the pc space and back to the original space ?
private Instances	m_transformedFormat The header for the transformed data format

Constructor Summary

PrincipalComponents()

Method Summary

private void	buildAttributeConstructor(Instances data)
void	buildEvaluator(Instances data) Initializes principal components and performs the analysis
Instance	convertInstance(Instance instance) Transform an instance in original (unormalized) format.
private Instance	convertInstanceToOriginal(Instance inst) Convert a pc transformed instance back to the original space
double	evaluateAttribute(int att) Evaluates the merit of a transformed attribute.
private void	fillCorrelation() Fill the correlation matrix
boolean	getNormalize() Gets whether or not input data is to be normalized
java.lang.String[]	getOptions() Gets the current settings of PrincipalComponents
boolean	getTransformBackToOriginal() Gets whether the data is to be transformed back to the original space.
double	getVarianceCovered() Gets the proportion of total variance to account for when retaining principal components
java.lang.String	globalInfo() Returns a string describing this attribute transformer
java.util.Enumeration	listOptions() Returns an enumeration describing the available options.
static void	main(java.lang.String[] argv) Main method for testing this class
private java.lang.String	matrixToString(double[][] matrix) Return a matrix as a String
java.lang.String	normalizeTipText() Returns the tip text for this property
private java.lang.String	principalComponentsSummary() Return a summary of the analysis
private void	resetOptions() Reset to defaults
void	setNormalize(boolean n) Set whether input data will be normalized.
void	setOptions(java.lang.String[] options) Parses a given list of options.
private Instances	setOutputFormat() Set the format for the transformed data
private Instances	setOutputFormatOriginal() Set up the header for the PC->original space dataset
void	setTransformBackToOriginal(boolean b) Sets whether the data should be transformed back to the original space
void	setVarianceCovered(double vc) Sets the amount of variance to account for when retaining principal components
java.lang.String	toString() Returns a description of this attribute transformer
java.lang.String	transformBackToOriginalTipText() Returns the tip text for this property
Instances	transformedData() Gets the transformed training data.
Instances	transformedHeader() Returns just the header for the transformed data (ie. an empty set of instances.
java.lang.String	varianceCoveredTipText() Returns the tip text for this property

Methods inherited from class weka.attributeSelection.ASEvaluation

forName, makeCopies, postProcess

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 data to transform analyse/transform

m_trainCopy

private Instances m_trainCopy

Keep a copy for the class attribute (if set)

m_transformedFormat

private Instances m_transformedFormat

The header for the transformed data format

m_originalSpaceFormat

private Instances m_originalSpaceFormat

The header for data transformed back to the original space

m_hasClass

private boolean m_hasClass

Data has a class set

m_classIndex

private int m_classIndex

Class index

m_numAttribs

private int m_numAttribs

Number of attributes

m_numInstances

private int m_numInstances

Number of instances

m_correlation

private double[][] m_correlation

Correlation matrix for the original data

m_eigenvectors

private double[][] m_eigenvectors

Will hold the unordered linear transformations of the (normalized) original data

m_eigenvalues

private double[] m_eigenvalues

Eigenvalues for the corresponding eigenvectors

m_sortedEigens

private int[] m_sortedEigens

Sorted eigenvalues

m_sumOfEigenValues

private double m_sumOfEigenValues

sum of the eigenvalues

m_replaceMissingFilter

private ReplaceMissingValues m_replaceMissingFilter

Filters for original data

m_normalizeFilter

private Normalize m_normalizeFilter

m_nominalToBinFilter

private NominalToBinary m_nominalToBinFilter

m_attributeFilter

private Remove m_attributeFilter

m_attribFilter

private Remove m_attribFilter

used to remove the class column if a class column is set

m_outputNumAtts

private int m_outputNumAtts

The number of attributes in the pc transformed data

m_normalize

private boolean m_normalize

normalize the input data?

m_coverVariance

private double m_coverVariance

the amount of varaince to cover in the original data when retaining the best n PC's

m_transBackToOriginal

private boolean m_transBackToOriginal

transform the data through the pc space and back to the original space ?

m_eTranspose

private double[][] m_eTranspose

holds the transposed eigenvectors for converting back to the original space

Constructor Detail

PrincipalComponents

public PrincipalComponents()

Method Detail

globalInfo

public java.lang.String globalInfo()

Returns a string describing this attribute transformer

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 a given list of options. Valid options are:

-N
Don't normalize the input data.

-R
Retain enough pcs to account for this proportion of the variance.

-T
Transform through the PC space and back to the original space.

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

resetOptions

private void resetOptions()

Reset to defaults

normalizeTipText

public java.lang.String normalizeTipText()

Returns the tip text for this property

Returns:

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

setNormalize

public void setNormalize(boolean n)

Set whether input data will be normalized.

Parameters:

n - true if input data is to be normalized

getNormalize

public boolean getNormalize()

Gets whether or not input data is to be normalized

Returns:

true if input data is to be normalized

varianceCoveredTipText

public java.lang.String varianceCoveredTipText()

Returns the tip text for this property

Returns:

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

setVarianceCovered

public void setVarianceCovered(double vc)

Sets the amount of variance to account for when retaining principal components

Parameters:

vc - the proportion of total variance to account for

getVarianceCovered

public double getVarianceCovered()

Gets the proportion of total variance to account for when retaining principal components

Returns:

the proportion of variance to account for

transformBackToOriginalTipText

public java.lang.String transformBackToOriginalTipText()

Returns the tip text for this property

Returns:

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

setTransformBackToOriginal

public void setTransformBackToOriginal(boolean b)

Sets whether the data should be transformed back to the original space

Parameters:

b - true if the data should be transformed back to the original space

getTransformBackToOriginal

public boolean getTransformBackToOriginal()

Gets whether the data is to be transformed back to the original space.

Returns:

true if the data is to be transformed back to the original space

getOptions

public java.lang.String[] getOptions()

Gets the current settings of PrincipalComponents

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

Initializes principal components and performs the analysis

Specified by:

buildEvaluator in class ASEvaluation

Parameters:

data - the instances to analyse/transform

Throws:

java.lang.Exception - if analysis fails

buildAttributeConstructor

private void buildAttributeConstructor(Instances data)
                                throws java.lang.Exception

Throws:

java.lang.Exception

transformedHeader

public Instances transformedHeader()
                            throws java.lang.Exception

Returns just the header for the transformed data (ie. an empty set of instances. This is so that AttributeSelection can determine the structure of the transformed data without actually having to get all the transformed data through getTransformedData().

Specified by:

transformedHeader in interface AttributeTransformer

Returns:

the header of the transformed data.

Throws:

java.lang.Exception - if the header of the transformed data can't be determined.

transformedData

public Instances transformedData()
                          throws java.lang.Exception

Gets the transformed training data.

Specified by:

transformedData in interface AttributeTransformer

Returns:

the transformed training data

Throws:

java.lang.Exception - if transformed data can't be returned

evaluateAttribute

public double evaluateAttribute(int att)
                         throws java.lang.Exception

Evaluates the merit of a transformed attribute. This is defined to be 1 minus the cumulative variance explained. Merit can't be meaningfully evaluated if the data is to be transformed back to the original space.

Specified by:

evaluateAttribute in class AttributeEvaluator

Parameters:

att - the attribute to be evaluated

Returns:

the merit of a transformed attribute

Throws:

java.lang.Exception - if attribute can't be evaluated

fillCorrelation

private void fillCorrelation()

Fill the correlation matrix

principalComponentsSummary

private java.lang.String principalComponentsSummary()

Return a summary of the analysis

Returns:

a summary of the analysis.

toString

public java.lang.String toString()

Returns a description of this attribute transformer

Returns:

a String describing this attribute transformer

matrixToString

private java.lang.String matrixToString(double[][] matrix)

Return a matrix as a String

Parameters:

matrix - that is decribed as a string

Returns:

a String describing a matrix

convertInstanceToOriginal

private Instance convertInstanceToOriginal(Instance inst)
                                    throws java.lang.Exception

Convert a pc transformed instance back to the original space

Throws:

java.lang.Exception

convertInstance

public Instance convertInstance(Instance instance)
                         throws java.lang.Exception

Transform an instance in original (unormalized) format. Convert back to the original space if requested.

Specified by:

convertInstance in interface AttributeTransformer

Parameters:

instance - an instance in the original (unormalized) format

Returns:

a transformed instance

Throws:

java.lang.Exception - if instance cant be transformed

setOutputFormatOriginal

private Instances setOutputFormatOriginal()
                                   throws java.lang.Exception

Set up the header for the PC->original space dataset

Throws:

java.lang.Exception

setOutputFormat

private Instances setOutputFormat()
                           throws java.lang.Exception

Set the format for the transformed data

Returns:

a set of empty Instances (header only) in the new format

Throws:

java.lang.Exception - if the output format can't be set

main

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

Main method for testing this class

Parameters:

argv - should contain the command line arguments to the evaluator/transformer (see AttributeSelection)