jsat.classifiers.imbalance

Class BorderlineSMOTE

java.lang.Object

jsat.classifiers.imbalance.SMOTE

jsat.classifiers.imbalance.BorderlineSMOTE

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, Classifier, Parameterized

public class BorderlineSMOTE
extends SMOTE

This class implements the Borderline extension of the SMOTE algorithm for dealing with class imbalance. SMOTE over-samples from the minority class at random points in the space. Borderline smote attempts to estimate which points are on the border of the class bounder, and over-samples only from the points on the boarder. Boarderline-SMOTE can also choose to perform interpolation using samples for the majority class, which can sometimes improve performance. The border is impacted by changes to the number of neighbors used. In the rare event that a boarder can't be estimated, this implementation will fall back to standard SMOTE.
This implementation extends the original SMOTE algorithm to the multi-class case.

See: Han, H., Wang, W.-Y., & Mao, B.-H. (2005). Borderline-SMOTE: A New Over-sampling Method in Imbalanced Data Sets Learning. In Proceedings of the 2005 International Conference on Advances in Intelligent Computing - Volume Part I (pp. 878–887). Berlin, Heidelberg: Springer-Verlag. DOI:10.1007/11538059_91

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
BorderlineSMOTE(BorderlineSMOTE toCopy) Copy constructor
BorderlineSMOTE(Classifier baseClassifier) Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.
BorderlineSMOTE(Classifier baseClassifier, boolean majorityInterpolation) Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.
BorderlineSMOTE(Classifier baseClassifier, DistanceMetric dm, boolean majorityInterpolation) Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.
BorderlineSMOTE(Classifier baseClassifier, DistanceMetric dm, double targetRatio, boolean majorityInterpolation) Creates a new Borderline-SMOTE model.
BorderlineSMOTE(Classifier baseClassifier, DistanceMetric dm, int smoteNeighbors, double targetRatio, boolean majorityInterpolation) Creates a new SMOTE object

Method Summary

Modifier and Type	Method and Description
BorderlineSMOTE	clone()
boolean	isMajorityInterpolation()
void	setMajorityInterpolation(boolean majorityInterpolation) Sets whether the generation of synthetic samples can make use of the majority samples (i.e., from other classes) or not.
void	train(ClassificationDataSet dataSet, boolean parallel) Trains the classifier and constructs a model for classification using the given data set.

Methods inherited from class jsat.classifiers.imbalance.SMOTE

classify, getBaseClassifier, getDistanceMetric, getSmoteNeighbors, getTargetRatio, setBaseClassifier, setDistanceMetric, setSmoteNeighbors, setTargetRatio, supportsWeightedData, train

Methods inherited from class java.lang.Object

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

Methods inherited from interface jsat.parameters.Parameterized

getParameter, getParameters

Constructor Detail

BorderlineSMOTE

public BorderlineSMOTE(Classifier baseClassifier)

Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class. It will not use majority interpolation.

Parameters:

baseClassifier - the base classifier to use after the SMOTEing is done.

BorderlineSMOTE

public BorderlineSMOTE(Classifier baseClassifier,
                       boolean majorityInterpolation)

Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.

Parameters:

baseClassifier - the base classifier to use after the SMOTEing is done.

majorityInterpolation - true if synthetic examples should use the majority class as well, or false to use only the minority class.

BorderlineSMOTE

public BorderlineSMOTE(Classifier baseClassifier,
                       DistanceMetric dm,
                       boolean majorityInterpolation)

Creates a new Borderline-SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.

Parameters:

baseClassifier - the base classifier to use after the SMOTEing is done.

dm - the distance metric to use for determining nearest neighbors

majorityInterpolation - true if synthetic examples should use the majority class as well, or false to use only the minority class.

BorderlineSMOTE

public BorderlineSMOTE(Classifier baseClassifier,
                       DistanceMetric dm,
                       double targetRatio,
                       boolean majorityInterpolation)

Creates a new Borderline-SMOTE model.

Parameters:

baseClassifier - the base classifier to use after the SMOTEing is done.

dm - the distance metric to use for determining nearest neighbors

targetRatio - the desired ratio of samples for each class with respect to the majority class.

majorityInterpolation - true if synthetic examples should use the majority class as well, or false to use only the minority class.

BorderlineSMOTE

public BorderlineSMOTE(Classifier baseClassifier,
                       DistanceMetric dm,
                       int smoteNeighbors,
                       double targetRatio,
                       boolean majorityInterpolation)

Creates a new SMOTE object

Parameters:

baseClassifier - the base classifier to use after the SMOTEing is done.

dm - the distance metric to use for determining nearest neighbors

smoteNeighbors - the number of neighbors to look at when interpolating points

targetRatio - the desired ratio of samples for each class with respect to the majority class.

majorityInterpolation - true if synthetic examples should use the majority class as well, or false to use only the minority class.

BorderlineSMOTE

public BorderlineSMOTE(BorderlineSMOTE toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

setMajorityInterpolation

public void setMajorityInterpolation(boolean majorityInterpolation)

Sets whether the generation of synthetic samples can make use of the majority samples (i.e., from other classes) or not. The use of majority samples is "Borderline-SMOTE2" in the original paper. If majority samples are not used, it is equivalent to "Borderline-SMOTE1".

Parameters:

majorityInterpolation - true if majority samples should be used for interpolation, and false if only minority samples should be used.

isMajorityInterpolation

public boolean isMajorityInterpolation()

Returns:

true if majority samples should be used for interpolation, and false if only minority samples should be used.

train

public void train(ClassificationDataSet dataSet,
                  boolean parallel)

Description copied from interface: Classifier

Trains the classifier and constructs a model for classification using the given data set. If the training method knows how, it will used the threadPool to conduct training in parallel. This method will block until the training has completed.

Specified by:

train in interface Classifier

Overrides:

train in class SMOTE

Parameters:

dataSet - the data set to train on

parallel - true if multiple threads should be used to train the model. false if it should be done in a single threaded manner.

clone

public BorderlineSMOTE clone()

Specified by:

clone in interface Classifier

Overrides:

clone in class SMOTE