Package unifeat.featureSelection.filter.supervised

Class MRMR

java.lang.Object

unifeat.featureSelection.FeatureSelection

unifeat.featureSelection.filter.FilterApproach

unifeat.featureSelection.filter.supervised.MRMR

public class MRMR
extends FilterApproach

This java class is used to implement the minimal redundancy maximal relevance (mRMR) method.

Author:

Sina Tabakhi

See Also:

• FilterApproach
• FeatureSelection

Field Summary

Fields

Modifier and Type	Field	Description
private double[][]	probFeature
private double[][]	valuesFeature

Fields inherited from class unifeat.featureSelection.FeatureSelection

numClass, numFeatures, numSelectedFeature, selectedFeatureSubset, trainSet

Constructor Summary

Constructors

Constructor	Description
MRMR(int sizeSelectedFeatureSubset)	Initializes the parameters
MRMR(Object... arguments)	Initializes the parameters

Method Summary

Modifier and Type	Method	Description
private void	computeDifferentValue(int index)	Saves the different values of each feature
private double[]	computeJointProb(int indexFeat2, int indexStartData, int indexEndData)	Computes the joint probabilities values between two features
private double	computeMutualInfo(int index1, int index2)	Computes the mutual information values between two features
private int	computeNumValue(int index)	Computes the number of different values of each feature
private void	computeProbFeat(int index)	Computes the probabilities values of each feature
void	evaluateFeatures()	Starts the feature selection process by minimal redundancy maximal relevance (mRMR) method
private int	findMaxWithIndex(double[] array)	Finds the maximum value in the array and returns its index
private boolean	isSelectedFeature(int index, int currentSize)	Checks that is the current feature (index) available in the subset of selected feature

Methods inherited from class unifeat.featureSelection.filter.FilterApproach

newMethod

Methods inherited from class unifeat.featureSelection.FeatureSelection

getSelectedFeatureSubset, loadDataSet, loadDataSet, setNumSelectedFeature, validate

Methods inherited from class java.lang.Object

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

Field Details

probFeature

private double[][] probFeature

valuesFeature

private double[][] valuesFeature

Constructor Details

MRMR

public MRMR(Object... arguments)

Initializes the parameters

Parameters:

arguments - array of parameters contains (sizeSelectedFeatureSubset) in which sizeSelectedFeatureSubset is the number of selected features

MRMR

public MRMR(int sizeSelectedFeatureSubset)

Initializes the parameters

Parameters:

sizeSelectedFeatureSubset - the number of selected features

Method Details

computeNumValue

private int computeNumValue(int index)

Computes the number of different values of each feature

Parameters:

index - the index of the feature

Returns:

the number of different values

computeDifferentValue

private void computeDifferentValue(int index)

Saves the different values of each feature

Parameters:

index - the index of the feature

computeProbFeat

private void computeProbFeat(int index)

Computes the probabilities values of each feature

Parameters:

index - the index of the feature

computeJointProb

private double[] computeJointProb(int indexFeat2,
 int indexStartData,
 int indexEndData)

Computes the joint probabilities values between two features

Parameters:

indexFeat2 - the index of the second feature

indexStartData - the start index of the dataset

indexEndData - the end index of the dataset

Returns:

an array of the joint probabilities values

computeMutualInfo

private double computeMutualInfo(int index1,
 int index2)

Computes the mutual information values between two features

Parameters:

index1 - the index of the first feature

index2 - the index of the second feature

Returns:

the mutual information value

findMaxWithIndex

private int findMaxWithIndex(double[] array)

Finds the maximum value in the array and returns its index

Parameters:

array - the input array

Returns:

the index of the maximum value in the array

isSelectedFeature

private boolean isSelectedFeature(int index,
 int currentSize)

Checks that is the current feature (index) available in the subset of selected feature

Parameters:

index - the index of the feature

currentSize - the current size of the selected features subset

Returns:

true if the current feature has been selected

evaluateFeatures

public void evaluateFeatures()

Starts the feature selection process by minimal redundancy maximal relevance (mRMR) method

Specified by:

evaluateFeatures in class FeatureSelection