meka.classifiers.multilabel

Class CC

java.lang.Object

weka.classifiers.AbstractClassifier

weka.classifiers.SingleClassifierEnhancer

meka.classifiers.multilabel.MultilabelClassifier

meka.classifiers.multilabel.CC

All Implemented Interfaces:

java.io.Serializable, java.lang.Cloneable, weka.classifiers.Classifier, weka.core.CapabilitiesHandler, weka.core.OptionHandler, weka.core.Randomizable, weka.core.RevisionHandler, weka.core.TechnicalInformationHandler

Direct Known Subclasses:

BCC, CC, CCp, CCUpdateable, MCC, PCC

public class CC
extends MultilabelClassifier
implements weka.core.Randomizable, weka.core.TechnicalInformationHandler

CC.java - The Classifier Chains Method. Like BR, but label outputs become new inputs for the next classifiers in the chain.
See: Jesse Read, Bernhard Pfahringer, Geoff Holmes, Eibe Frank. Classifier Chains for Multi-label Classification. Machine Learning Journal. Springer. Vol. 85(3), pp 333-359. (May 2011).
See: Jesse Read, Bernhard Pfahringer, Geoff Holmes, Eibe Frank. Classifier Chains for Multi-label Classification. In Proc. of 20th European Conference on Machine Learning (ECML 2009). Bled, Slovenia, September 2009.
Note that the code was reorganised substantially since earlier versions, to accomodate additional functionality needed for e.g., MCC, PCC.

Version:

December 2013

Author:

Jesse Read

See Also:

Serialized Form

Constructor Summary

Constructors

Constructor and Description
CC()

Method Summary

Methods

Modifier and Type	Method and Description
void	buildClassifier(weka.core.Instances D)
double[]	distributionForInstance(weka.core.Instance x)
int[]	getChain()
double[]	getConfidences() GetConfidences - get the posterior probabilities of the previous prediction (after calling distributionForInstance(x)).
int	getSeed()
weka.core.TechnicalInformation	getTechnicalInformation()
weka.core.Instance[]	getTransformTemplates(weka.core.Instance x) GetTransformTemplates - pre-transform the instance x, to make things faster.
java.lang.String	globalInfo() Description to display in the GUI.
static void	main(java.lang.String[] args)
double[]	probabilityForInstance(weka.core.Instance x, double[] path) ProbabilityForInstance - Force our way down the imposed 'path'.
void	rebuildClassifier(int[] new_chain, weka.core.Instances D) Rebuild - NOT YET IMPLEMENTED.
double[]	sampleForInstance(weka.core.Instance x, java.util.Random r) SampleForInstance.
double[]	sampleForInstanceFast(weka.core.Instance[] t_, java.util.Random r) SampleForInstance - given an Instance template for each label, and a Random.
void	setChain(int[] chain)
void	setSeed(int s)
weka.core.Instance[]	transformInstance(weka.core.Instance x) TransformInstances - this function is DEPRECATED.

Methods inherited from class meka.classifiers.multilabel.MultilabelClassifier

evaluation, getCapabilities, getRevision, getTemplate, makeCopies, runClassifier, testCapabilities, toString

Methods inherited from class weka.classifiers.SingleClassifierEnhancer

classifierTipText, getClassifier, getOptions, listOptions, setClassifier, setOptions

Methods inherited from class weka.classifiers.AbstractClassifier

classifyInstance, debugTipText, forName, getDebug, makeCopies, makeCopy, runClassifier, setDebug

Methods inherited from class java.lang.Object

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

Constructor Detail

CC

public CC()

Method Detail

setChain

public void setChain(int[] chain)

getChain

public int[] getChain()

buildClassifier

public void buildClassifier(weka.core.Instances D)
                     throws java.lang.Exception

Specified by:

buildClassifier in interface weka.classifiers.Classifier

Specified by:

buildClassifier in class MultilabelClassifier

Throws:

java.lang.Exception

getConfidences

public double[] getConfidences()

GetConfidences - get the posterior probabilities of the previous prediction (after calling distributionForInstance(x)).

distributionForInstance

public double[] distributionForInstance(weka.core.Instance x)
                                 throws java.lang.Exception

Specified by:

distributionForInstance in interface weka.classifiers.Classifier

Specified by:

distributionForInstance in class MultilabelClassifier

Throws:

java.lang.Exception

sampleForInstance

public double[] sampleForInstance(weka.core.Instance x,
                         java.util.Random r)
                           throws java.lang.Exception

SampleForInstance. predict y[j] stochastically rather than deterministically (as with distributionForInstance(Instance x)).

Parameters:

x - test Instance

r - Random <- TODO probably can use this.m_R instead

Throws:

java.lang.Exception

getTransformTemplates

public weka.core.Instance[] getTransformTemplates(weka.core.Instance x)
                                           throws java.lang.Exception

GetTransformTemplates - pre-transform the instance x, to make things faster.

Returns:

the templates

Throws:

java.lang.Exception

sampleForInstanceFast

public double[] sampleForInstanceFast(weka.core.Instance[] t_,
                             java.util.Random r)
                               throws java.lang.Exception

SampleForInstance - given an Instance template for each label, and a Random.

Parameters:

t_ - Instance templates (pre-transformed) using #getTransformTemplates(x)

Throws:

java.lang.Exception

transformInstance

public weka.core.Instance[] transformInstance(weka.core.Instance x)
                                       throws java.lang.Exception

TransformInstances - this function is DEPRECATED. this function preloads the instances with the correct class labels ... to make the chain much faster, but CNode does not yet have this functionality ... need to do something about this!

Throws:

java.lang.Exception

probabilityForInstance

public double[] probabilityForInstance(weka.core.Instance x,
                              double[] path)
                                throws java.lang.Exception

ProbabilityForInstance - Force our way down the imposed 'path'.
TODO rename distributionForPath ? and simplify like distributionForInstance ?
For example p (y=1010|x) = [0.9,0.8,0.1,0.2]. If the product = 1, this is probably the correct path!

Parameters:

x - test Instance

path - the path we want to go down

Returns:

the probabilities associated with this path: [p(Y_1==path[1]|x),...,p(Y_L==path[L]|x)]

Throws:

java.lang.Exception

rebuildClassifier

public void rebuildClassifier(int[] new_chain,
                     weka.core.Instances D)
                       throws java.lang.Exception

Rebuild - NOT YET IMPLEMENTED. For efficiency reasons, we may want to rebuild part of the chain. If chain[] = [1,2,3,4] and new_chain[] = [1,2,4,3] we only need to rebuild the final two links.

Parameters:

new_chain - the new chain

D - the original training data

Throws:

java.lang.Exception

setSeed

public void setSeed(int s)

Specified by:

setSeed in interface weka.core.Randomizable

getSeed

public int getSeed()

Specified by:

getSeed in interface weka.core.Randomizable

globalInfo

public java.lang.String globalInfo()

Description to display in the GUI.

Overrides:

globalInfo in class MultilabelClassifier

Returns:

the description

getTechnicalInformation

public weka.core.TechnicalInformation getTechnicalInformation()

Specified by:

getTechnicalInformation in interface weka.core.TechnicalInformationHandler

main

public static void main(java.lang.String[] args)