jsat.datatransform.featureselection

Class ReliefF

java.lang.Object

jsat.datatransform.RemoveAttributeTransform

jsat.datatransform.featureselection.ReliefF

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, DataTransform

public class ReliefF
extends RemoveAttributeTransform

Provides an implementation of the ReliefF algorithm for feature importance computing. Because JSAT does not support neighbor searching for categorical values, it does not provides weights for categorical variables.
Weight values are in the range [-1, 1]. The value is a measure of corelation, so the absolute value of the individual weights would form its relative importance to the others.
The ReliefF algorithm is meant for classification problems, and is computed in a nearest neighbor fashion.

See:
Kononenko, I., Simec, E.,&Robnik-Sikonja, M. (1997). Overcoming the myopia of inductive learning algorithms with RELIEFF. Applied Intelligence, 7, 39–55.

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
ReliefF(ClassificationDataSet cds, int featureCount, int m, int n, DistanceMetric dm) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(ClassificationDataSet cds, int featureCount, int m, int n, DistanceMetric dm, java.util.concurrent.ExecutorService threadPool) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(ClassificationDataSet cds, int featureCount, int m, int n, DistanceMetric dm, VectorCollection<Vec> vc) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(ClassificationDataSet cds, int featureCount, int m, int n, DistanceMetric dm, VectorCollection<Vec> vcf, java.util.concurrent.ExecutorService threadPool) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(int featureCount) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(int featureCount, int m, int n, DistanceMetric dm) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.
ReliefF(int featureCount, int m, int n, DistanceMetric dm, VectorCollection<Vec> vc) Creates a new ReliefF object to measure the importance of the variables with respect to a classification task.

Method Summary

Modifier and Type	Method and Description
ReliefF	clone()
void	fit(DataSet data) Fits this transform to the given dataset.
void	fit(DataSet data, java.util.concurrent.ExecutorService threadPool)
DistanceMetric	getDistanceMetric() Returns the distance metric to use
int	getFeatureCount() Returns the number of features to sue
int	getIterations() Returns the number of iterations to use
int	getNeighbors() Returns the number of neighbors that will be used at each step of the algorithm.
Vec	getWeights() Returns accesses to the learned weight data.
void	setDistanceMetric(DistanceMetric dm) Sets the distance metric to infer the feature importance with
void	setFeatureCount(int featureCount) Sets the number of features to select for use from the set of all input features
void	setIterations(int iterations) Sets the number of iterations of the ReliefF algorithm that will be run
void	setNeighbors(int neighbors) Sets the number of neighbors to use to infer feature importance from

Methods inherited from class jsat.datatransform.RemoveAttributeTransform

consolidate, getKeptNominal, getKeptNumeric, getReverseNominalMap, getReverseNumericMap, transform

Methods inherited from class java.lang.Object

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

Constructor Detail

ReliefF

public ReliefF(int featureCount)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

featureCount - the number of features to keep

ReliefF

public ReliefF(int featureCount,
               int m,
               int n,
               DistanceMetric dm)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

ReliefF

public ReliefF(ClassificationDataSet cds,
               int featureCount,
               int m,
               int n,
               DistanceMetric dm)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

cds - the data set to measure numeric variable importance from

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

ReliefF

public ReliefF(ClassificationDataSet cds,
               int featureCount,
               int m,
               int n,
               DistanceMetric dm,
               java.util.concurrent.ExecutorService threadPool)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

cds - the data set to measure numeric variable importance from

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

threadPool - the source of threads to use for the computation

ReliefF

public ReliefF(ClassificationDataSet cds,
               int featureCount,
               int m,
               int n,
               DistanceMetric dm,
               VectorCollection<Vec> vc)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

cds - the data set to measure numeric variable importance from

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

vc - the vector collection to create accelerating structures for nearest neighbor

ReliefF

public ReliefF(int featureCount,
               int m,
               int n,
               DistanceMetric dm,
               VectorCollection<Vec> vc)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

vc - the factor to create accelerating structures for nearest neighbor

ReliefF

public ReliefF(ClassificationDataSet cds,
               int featureCount,
               int m,
               int n,
               DistanceMetric dm,
               VectorCollection<Vec> vcf,
               java.util.concurrent.ExecutorService threadPool)

Creates a new ReliefF object to measure the importance of the variables with respect to a classification task. Only numeric features will be removed. Categorical features will be ignored and left in tact by the transformation

Parameters:

cds - the data set to measure numeric variable importance from

featureCount - the number of features to keep

m - the number of learning iterations to perform

n - the number of neighbors to measure importance from

dm - the distance metric to use

vcf - the factor to create accelerating structures for nearest neighbor

threadPool - the source of threads to use for the computation

Method Detail

fit

public void fit(DataSet data)

Description copied from interface: DataTransform

Fits this transform to the given dataset. Some transforms can only be learned from classification or regression datasets. If an incompatible dataset type is given, a FailedToFitException exception may be thrown.

Specified by:

fit in interface DataTransform

Overrides:

fit in class RemoveAttributeTransform

Parameters:

data - the dataset to fir this transform to

fit

public void fit(DataSet data,
                java.util.concurrent.ExecutorService threadPool)

getWeights

public Vec getWeights()

Returns accesses to the learned weight data. Altering the values will be reflected in this original ReliefF object.

Returns:

access to the raw weight values

clone

public ReliefF clone()

Specified by:

clone in interface DataTransform

Overrides:

clone in class RemoveAttributeTransform

setFeatureCount

public void setFeatureCount(int featureCount)

Sets the number of features to select for use from the set of all input features

Parameters:

featureCount - the number of features to use

getFeatureCount

public int getFeatureCount()

Returns the number of features to sue

Returns:

the number of features to sue

setIterations

public void setIterations(int iterations)

Sets the number of iterations of the ReliefF algorithm that will be run

Parameters:

iterations - the number of iterations to run

getIterations

public int getIterations()

Returns the number of iterations to use

Returns:

the number of iterations to use

setNeighbors

public void setNeighbors(int neighbors)

Sets the number of neighbors to use to infer feature importance from

Parameters:

neighbors - the number of neighbors to use

getNeighbors

public int getNeighbors()

Returns the number of neighbors that will be used at each step of the algorithm.

Returns:

the number of neighbors that will be used

setDistanceMetric

public void setDistanceMetric(DistanceMetric dm)

Sets the distance metric to infer the feature importance with

Parameters:

dm - the distance metric to use

getDistanceMetric

public DistanceMetric getDistanceMetric()

Returns the distance metric to use

Returns:

the distance metric to use