Package moa.classifiers.meta.imbalanced

Class OnlineAdaC2

java.lang.Object

moa.AbstractMOAObject

moa.options.AbstractOptionHandler

moa.classifiers.AbstractClassifier

moa.classifiers.meta.imbalanced.OnlineAdaC2

All Implemented Interfaces:

Configurable, java.io.Serializable, CapabilitiesHandler, Classifier, MultiClassClassifier, AWTRenderable, Learner<Example<Instance>>, MOAObject, OptionHandler

public class OnlineAdaC2
extends AbstractClassifier
implements MultiClassClassifier, CapabilitiesHandler

OnlineAdaC2 is the adaptation of the ensemble learner to data streams

AdaC2 is a boosting algorithm that takes the different misclassification costs into consideration when calculating the classier weights, and updates the sample weight. AdaC2 increases more the weights on the misclassified positive samples, compared to misclassified negative samples.:math:`C_P` and :math:`C_N` are the positive and negative costs.

This online ensemble learner method is improved by the addition of an ADWIN change detector. ADWIN stands for Adaptive Windowing. It works by keeping updated statistics of a variable sized window, so it can detect changes and perform cuts in its window to better adapt the learning algorithms.

See details in:
B. Wang and J. Pineau, "Online Bagging and Boosting for Imbalanced Data Streams," in IEEE Transactions on Knowledge and Data Engineering, vol. 28, no. 12, pp. 3353-3366, 1 Dec. 2016. doi: 10.1109/TKDE.2016.2609424

Parameters:

• -l : Each classifier to train of the ensemble is an instance of the base estimator.
• -s : The size of the ensemble, in other words, how many classifiers to train.
• -p : The cost of misclassifying a positive sample.
• -n : The cost of misclassifying a negative sample.
• -d : Should use ADWIN as drift detector? If enabled it is used by the method to track the performance of the classifiers and adapt when a drift is detected.
• -r : Seed for the random state.

Version:

$Revision: 1 $

Author:

Alessio Bernardo (alessio dot bernardo at polimi dot dot it)

See Also:

Serialized Form

Field Summary

Fields

Modifier and Type	Field	Description
protected java.util.ArrayList<ADWIN>	adwinEnsemble
protected Classifier	baseLearner
ClassOption	baseLearnerOption
protected double	costNegative
FloatOption	costNegativeOption
protected double	costPositive
FloatOption	costPositiveOption
FlagOption	disableDriftDetectionOption
protected boolean	driftDetection
protected java.util.ArrayList<Classifier>	ensemble
IntOption	ensembleSizeOption
protected java.util.ArrayList<java.lang.Double>	lambdaFN
protected java.util.ArrayList<java.lang.Double>	lambdaFP
protected java.util.ArrayList<java.lang.Double>	lambdaSum
protected java.util.ArrayList<java.lang.Double>	lambdaTN
protected java.util.ArrayList<java.lang.Double>	lambdaTP
protected int	nEstimators
protected java.util.ArrayList<java.lang.Double>	wAcc
protected java.util.ArrayList<java.lang.Double>	wErr

Fields inherited from class moa.classifiers.AbstractClassifier

classifierRandom, modelContext, randomSeed, randomSeedOption, trainingWeightSeenByModel

Fields inherited from class moa.options.AbstractOptionHandler

config

Constructor Summary

Constructors

Constructor	Description
OnlineAdaC2()

Method Summary

Modifier and Type	Method	Description
protected void	adjustEnsembleSize(int nClasses)
ImmutableCapabilities	defineImmutableCapabilities()	Defines the set of capabilities the object has.
void	getModelDescription(java.lang.StringBuilder arg0, int arg1)	Returns a string representation of the model.
protected Measurement[]	getModelMeasurementsImpl()	Gets the current measurements of this classifier. The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods.
java.lang.String	getPurposeString()	Dictionary with option texts and objects
double[]	getVotesForInstance(Instance instance)	Predicts the class memberships for a given instance.
boolean	isRandomizable()	Gets whether this learner needs a random seed.
void	resetLearningImpl()	Resets this classifier.
void	trainOnInstanceImpl(Instance instance)	Trains this classifier incrementally using the given instance. The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods.

Methods inherited from class moa.classifiers.AbstractClassifier

contextIsCompatible, copy, correctlyClassifies, getAttributeNameString, getAWTRenderer, getClassLabelString, getClassNameString, getDescription, getModel, getModelContext, getModelMeasurements, getNominalValueString, getPredictionForInstance, getPredictionForInstance, getSubClassifiers, getSublearners, getVotesForInstance, modelAttIndexToInstanceAttIndex, modelAttIndexToInstanceAttIndex, prepareForUseImpl, resetLearning, setModelContext, setRandomSeed, trainingHasStarted, trainingWeightSeenByModel, trainOnInstance, trainOnInstance

Methods inherited from class moa.AbstractMOAObject

copy, measureByteSize, measureByteSize, toString

Methods inherited from class moa.options.AbstractOptionHandler

getCLICreationString, getOptions, getPreparedClassOption, prepareClassOptions, prepareForUse, prepareForUse

Methods inherited from interface moa.capabilities.CapabilitiesHandler

getCapabilities

Methods inherited from interface moa.MOAObject

measureByteSize

Methods inherited from class java.lang.Object

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

Methods inherited from interface moa.options.OptionHandler

getCLICreationString, getOptions, prepareForUse, prepareForUse

Field Detail

baseLearnerOption

public ClassOption baseLearnerOption

ensembleSizeOption

public IntOption ensembleSizeOption

costPositiveOption

public FloatOption costPositiveOption

costNegativeOption

public FloatOption costNegativeOption

disableDriftDetectionOption

public FlagOption disableDriftDetectionOption

baseLearner

protected Classifier baseLearner

nEstimators

protected int nEstimators

costPositive

protected double costPositive

costNegative

protected double costNegative

driftDetection

protected boolean driftDetection

ensemble

protected java.util.ArrayList<Classifier> ensemble

adwinEnsemble

protected java.util.ArrayList<ADWIN> adwinEnsemble

lambdaTP

protected java.util.ArrayList<java.lang.Double> lambdaTP

lambdaTN

protected java.util.ArrayList<java.lang.Double> lambdaTN

lambdaFP

protected java.util.ArrayList<java.lang.Double> lambdaFP

lambdaFN

protected java.util.ArrayList<java.lang.Double> lambdaFN

lambdaSum

protected java.util.ArrayList<java.lang.Double> lambdaSum

wAcc

protected java.util.ArrayList<java.lang.Double> wAcc

wErr

protected java.util.ArrayList<java.lang.Double> wErr

Constructor Detail

OnlineAdaC2

public OnlineAdaC2()

Method Detail

getPurposeString

public java.lang.String getPurposeString()

Description copied from class: AbstractOptionHandler

Dictionary with option texts and objects

Specified by:

getPurposeString in interface OptionHandler

Overrides:

getPurposeString in class AbstractClassifier

Returns:

the string with the purpose of this object

resetLearningImpl

public void resetLearningImpl()

Description copied from class: AbstractClassifier

Resets this classifier. It must be similar to starting a new classifier from scratch.

The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods. Note that this will produce compiler errors if not overridden.

Specified by:

resetLearningImpl in class AbstractClassifier

trainOnInstanceImpl

public void trainOnInstanceImpl(Instance instance)

Description copied from class: AbstractClassifier

Trains this classifier incrementally using the given instance.

The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods. Note that this will produce compiler errors if not overridden.

Specified by:

trainOnInstanceImpl in class AbstractClassifier

Parameters:

instance - the instance to be used for training

getVotesForInstance

public double[] getVotesForInstance(Instance instance)

Description copied from interface: Classifier

Predicts the class memberships for a given instance. If an instance is unclassified, the returned array elements must be all zero.

Specified by:

getVotesForInstance in interface Classifier

Specified by:

getVotesForInstance in class AbstractClassifier

Parameters:

instance - the instance to be classified

Returns:

an array containing the estimated membership probabilities of the test instance in each class

isRandomizable

public boolean isRandomizable()

Description copied from interface: Learner

Gets whether this learner needs a random seed. Examples of methods that needs a random seed are bagging and boosting.

Specified by:

isRandomizable in interface Learner<Example<Instance>>

Returns:

true if the learner needs a random seed.

getModelDescription

public void getModelDescription(java.lang.StringBuilder arg0,
                                int arg1)

Description copied from class: AbstractClassifier

Returns a string representation of the model.

Specified by:

getModelDescription in class AbstractClassifier

Parameters:

arg0 - the stringbuilder to add the description

arg1 - the number of characters to indent

getModelMeasurementsImpl

protected Measurement[] getModelMeasurementsImpl()

Description copied from class: AbstractClassifier

Gets the current measurements of this classifier.

The reason for ...Impl methods: ease programmer burden by not requiring them to remember calls to super in overridden methods. Note that this will produce compiler errors if not overridden.

Specified by:

getModelMeasurementsImpl in class AbstractClassifier

Returns:

an array of measurements to be used in evaluation tasks

adjustEnsembleSize

protected void adjustEnsembleSize(int nClasses)

defineImmutableCapabilities

public ImmutableCapabilities defineImmutableCapabilities()

Description copied from interface: CapabilitiesHandler

Defines the set of capabilities the object has. Should be overridden if the object's capabilities do not change.

Specified by:

defineImmutableCapabilities in interface CapabilitiesHandler

Overrides:

defineImmutableCapabilities in class AbstractClassifier

Returns:

The capabilities of the object.