jsat.classifiers.imbalance

Class SMOTE

java.lang.Object

jsat.classifiers.imbalance.SMOTE

All Implemented Interfaces:

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

Direct Known Subclasses:

BorderlineSMOTE

public class SMOTE
extends java.lang.Object
implements Classifier, Parameterized

This class implements the Synthetic Minority Over-sampling TEchnique (SMOTE) for dealing with class imbalance. It does this by over-sampling the minority classes to bring their total count up to parity (or some target ratio) with the majority class. This is done by interpolating between minority points and their neighbors to create new synthetic points that are not present in the current dataset. For this reason SMOTE only works with numeric feature vectors.

See: Chawla, N., Bowyer, K., Hall, L., & Kegelmeyer, P. (2002). SMOTE: synthetic minority over-sampling technique. Artificial Intelligence Research, 16, 321–357. Retrieved from here

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
SMOTE(Classifier baseClassifier) Creates a new SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class. This implementation extends the original SMOTE algorithm to the multi-class case.
SMOTE(Classifier baseClassifier, DistanceMetric dm) Creates a new SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.
SMOTE(Classifier baseClassifier, DistanceMetric dm, double targetRatio) Creates a new SMOTE model.
SMOTE(Classifier baseClassifier, DistanceMetric dm, int smoteNeighbors, double targetRatio) Creates a new SMOTE object
SMOTE(SMOTE toCopy) Copy constructor

Method Summary

Modifier and Type	Method and Description
CategoricalResults	classify(DataPoint data) Performs classification on the given data point.
SMOTE	clone()
Classifier	getBaseClassifier()
DistanceMetric	getDistanceMetric()
int	getSmoteNeighbors()
double	getTargetRatio()
void	setBaseClassifier(Classifier baseClassifier) Sets the classifier to use after the dataset has been modified
void	setDistanceMetric(DistanceMetric dm) Sets the metric used to determine the nearest neighbors of each point.
void	setSmoteNeighbors(int smoteNeighbors) Sets the number of neighbors that will be used to be candidates for interpolation.
void	setTargetRatio(double targetRatio) Sets the desired ratio of samples for each class compared to the majority class.
boolean	supportsWeightedData() Indicates whether the model knows how to train using weighted data points.
void	train(ClassificationDataSet dataSet) Trains the classifier and constructs a model for classification using the given data set.
void	train(ClassificationDataSet dataSet, boolean parallel) Trains the classifier and constructs a model for classification using the given data set.

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

SMOTE

public SMOTE(Classifier baseClassifier)

Creates a new SMOTE model that will over-sample the minority classes so that there is a balanced number of data points in each class.
This implementation extends the original SMOTE algorithm to the multi-class case.

Parameters:

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

SMOTE

public SMOTE(Classifier baseClassifier,
             DistanceMetric dm)

Creates a new 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

SMOTE

public SMOTE(Classifier baseClassifier,
             DistanceMetric dm,
             double targetRatio)

Creates a new 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.

SMOTE

public SMOTE(Classifier baseClassifier,
             DistanceMetric dm,
             int smoteNeighbors,
             double targetRatio)

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.

SMOTE

public SMOTE(SMOTE toCopy)

Copy constructor

Parameters:

toCopy - the object to copy

Method Detail

setDistanceMetric

public void setDistanceMetric(DistanceMetric dm)

Sets the metric used to determine the nearest neighbors of each point.

Parameters:

dm - the distance metric to use.

getDistanceMetric

public DistanceMetric getDistanceMetric()

Returns:

the distance metric to use

setSmoteNeighbors

public void setSmoteNeighbors(int smoteNeighbors)

Sets the number of neighbors that will be used to be candidates for interpolation. The default value recommended in the original paper is 5.

Parameters:

smoteNeighbors - the number of candidate neighbors to select from when creating synthetic data points.

getSmoteNeighbors

public int getSmoteNeighbors()

Returns:

the number of candidate neighbors to select from when creating synthetic data points.

setTargetRatio

public void setTargetRatio(double targetRatio)

Sets the desired ratio of samples for each class compared to the majority class. A ratio of 1.0 will oversample the minority classes until they have just as many data points as the majority class. If any minority class already exists at a ratio equal to or above this ratio, no over samples will be created for that class. If the target ratio is greater than one, all classes including the majority class will be over-sampled to the desired ratio.

Parameters:

targetRatio - the target ratio between each class and the majority class

getTargetRatio

public double getTargetRatio()

Returns:

the target ratio between each class and the majority class

setBaseClassifier

public void setBaseClassifier(Classifier baseClassifier)

Sets the classifier to use after the dataset has been modified

Parameters:

baseClassifier - the classifier to use for training and prediction

getBaseClassifier

public Classifier getBaseClassifier()

Returns:

the classifier used by the model

classify

public CategoricalResults classify(DataPoint data)

Description copied from interface: Classifier

Performs classification on the given data point.

Specified by:

classify in interface Classifier

Parameters:

data - the data point to classify

Returns:

the results of the classification.

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

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.

train

public void train(ClassificationDataSet dataSet)

Description copied from interface: Classifier

Trains the classifier and constructs a model for classification using the given data set.

Specified by:

train in interface Classifier

Parameters:

dataSet - the data set to train on

supportsWeightedData

public boolean supportsWeightedData()

Description copied from interface: Classifier

Indicates whether the model knows how to train using weighted data points. If it does, the model will train assuming the weights. The values returned by this method may change depending on the parameters set for the model.

Specified by:

supportsWeightedData in interface Classifier

Returns:

true if the model supports weighted data, false otherwise

clone

public SMOTE clone()

Specified by:

clone in interface Classifier

Overrides:

clone in class java.lang.Object