com.rapidminer.operator.learner.tree.criterions
Class AccuracyCriterion

java.lang.Object
  com.rapidminer.operator.learner.tree.criterions.AbstractCriterion
      com.rapidminer.operator.learner.tree.criterions.AccuracyCriterion

All Implemented Interfaces:

Criterion

public class AccuracyCriterion
extends AbstractCriterion

Calculates the accuracies for the given split if the children predict the majority classes.

Author:

Ingo Mierswa

Field Summary

Fields inherited from class com.rapidminer.operator.learner.tree.criterions.AbstractCriterion

labelAttribute, leftLabelWeights, leftWeight, rightLabelWeights, rightWeight, totalLabelWeights, totalWeight, weightAttribute

Constructor Summary

AccuracyCriterion()

Method Summary

double	getBenefit(double[][] weightCounts) This method will return the calculated benefit if the weights would have distributed over the labels as given.
double	getIncrementalBenefit()
double	getNominalBenefit(ExampleSet exampleSet, Attribute attribute)
double	getNumericalBenefit(ExampleSet exampleSet, Attribute attribute, double splitValue)
boolean	supportsIncrementalCalculation()

Methods inherited from class com.rapidminer.operator.learner.tree.criterions.AbstractCriterion

createCriterion, startIncrementalCalculation, swapExample

Methods inherited from class java.lang.Object

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

Constructor Detail

AccuracyCriterion

public AccuracyCriterion()

Method Detail

getNominalBenefit

public double getNominalBenefit(ExampleSet exampleSet,
                                Attribute attribute)

getNumericalBenefit

public double getNumericalBenefit(ExampleSet exampleSet,
                                  Attribute attribute,
                                  double splitValue)

getBenefit

public double getBenefit(double[][] weightCounts)

Description copied from interface: Criterion

This method will return the calculated benefit if the weights would have distributed over the labels as given. The first index specifies the split fraction, the second index the label. Henve the first index is always between 0 and 1 included for numerical splits, since only two sides can occur there. For splits on nominal attributes, the number of split sides is determined by the number of possible values.

supportsIncrementalCalculation

public boolean supportsIncrementalCalculation()

Specified by:

supportsIncrementalCalculation in interface Criterion

Overrides:

supportsIncrementalCalculation in class AbstractCriterion

getIncrementalBenefit

public double getIncrementalBenefit()

Specified by:

getIncrementalBenefit in interface Criterion

Overrides:

getIncrementalBenefit in class AbstractCriterion