de.jstacs.classifier
Class ScoreBasedPerformanceMeasureDefinitions

java.lang.Object
  de.jstacs.classifier.ScoreBasedPerformanceMeasureDefinitions

public class ScoreBasedPerformanceMeasureDefinitions
extends Object

This class contains the methods that are needed to evaluate a score based 2-class-classifier.

Author:

Jens Keilwagen, Jan Grau

See Also:

AbstractScoreBasedClassifier, AbstractClassifier.evaluate(MeasureParameters, boolean, de.jstacs.data.Sample[]), AbstractClassifier.evaluateAll(MeasureParameters, boolean, de.jstacs.data.Sample[]), MeasureParameters

Nested Class Summary

static class

ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair This class is used as a container that allows to store a threshold and the result of measure together.

Constructor Summary

ScoreBasedPerformanceMeasureDefinitions()

Method Summary

protected static double	getAUC_PR(double[] scoresClass0, double[] scoresClass1, ArrayList<double[]> list) This method computes the area under the precision recall curve.
protected static double	getAUC_ROC(double[] scoresClass0, double[] scoresClass1, ArrayList<double[]> list) This method computes the area under the receiver operator characteristics curve.
protected static double	getClassificationRateFor2Classes(double[] scoresClass0, double[] scoresClass1) This method computes the classification rate.
protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair	getFPRForSensitivity(double[] scoresClass0, double[] scoresClass1, double sensitivity) This method computes the false positive rate (FPR) for a given sensitivity.
protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair	getMaxOfCC(double[] scoresClass0, double[] scoresClass1) This method computes the maximal correlation coefficient (CC_max).
protected static double[][]	getPartialROC(double[] scoresClass0, double[] scoresClass1, RangeParameter specs) This method allows to compute are partial ROC curve.
protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair	getPPVForSensitivity(double[] scoresClass0, double[] scoresClass1, double sensitivity) This method computes the positive predictive value (PPV) for a given sensitivity.
protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair	getSensitivityForSpecificity(double[] scoresClass0, double[] scoresClass1, double specificity) This method computes the sensitivity for a given specificity.

Methods inherited from class java.lang.Object

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

Constructor Detail

ScoreBasedPerformanceMeasureDefinitions

public ScoreBasedPerformanceMeasureDefinitions()

Method Detail

getPartialROC

protected static double[][] getPartialROC(double[] scoresClass0,
                                          double[] scoresClass1,
                                          RangeParameter specs)
                                   throws ParameterException

This method allows to compute are partial ROC curve.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

specs - these specificities are used as grid for the partial curve

Returns:

a matrix

• m[0][i] i-th specificity
• m[1][i] i-th sensitivity
• m[2][i] i-th threshold

Throws:

ParameterException - if the method RangeParameter.next() throws one

getAUC_PR

protected static double getAUC_PR(double[] scoresClass0,
                                  double[] scoresClass1,
                                  ArrayList<double[]> list)

This method computes the area under the precision recall curve. It also enables the user to compute the complete curve so that it can be used for plotting. If you like to get the curve, you should use an empty ArrayList for list, otherwise use null.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

list - null or an empty list

Returns:

the area under the precision recall curve

getAUC_ROC

protected static double getAUC_ROC(double[] scoresClass0,
                                   double[] scoresClass1,
                                   ArrayList<double[]> list)

This method computes the area under the receiver operator characteristics curve. It also enables the user to compute the complete curve so that it can be used for plotting. If you like to get the curve, you should use an empty ArrayList for list, otherwise use null.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

list - null or an empty list

Returns:

the area under the receiver operator characteristics curve

getFPRForSensitivity

protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair getFPRForSensitivity(double[] scoresClass0,
                                                                                                   double[] scoresClass1,
                                                                                                   double sensitivity)
                                                                                            throws IllegalArgumentException

This method computes the false positive rate (FPR) for a given sensitivity.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

sensitivity - the fixed sensitivity which is used to measure the FPR

Returns:

the pair containing the threshold and the FPR

Throws:

IllegalArgumentException - if the sensitivity is not in (0,1)

getPPVForSensitivity

protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair getPPVForSensitivity(double[] scoresClass0,
                                                                                                   double[] scoresClass1,
                                                                                                   double sensitivity)
                                                                                            throws IllegalArgumentException

This method computes the positive predictive value (PPV) for a given sensitivity.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

sensitivity - the fixed sensitivity which is used to measure the PPV

Returns:

the pair containing the threshold and the PPV

Throws:

IllegalArgumentException - if the sensitivity is not in (0,1)

getMaxOfCC

protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair getMaxOfCC(double[] scoresClass0,
                                                                                         double[] scoresClass1)
                                                                                  throws Exception

This method computes the maximal correlation coefficient (CC_max).

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

Returns:

the pair containing the threshold and the CC_max

Throws:

Exception - if the CC_max can not be computed

getSensitivityForSpecificity

protected static ScoreBasedPerformanceMeasureDefinitions.ThresholdMeasurePair getSensitivityForSpecificity(double[] scoresClass0,
                                                                                                           double[] scoresClass1,
                                                                                                           double specificity)
                                                                                                    throws IllegalArgumentException

This method computes the sensitivity for a given specificity.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

specificity - the specificity that is used to measure the sensitivity

Returns:

the pair containing the threshold and the sensitivity

Throws:

IllegalArgumentException - if the specificity is not in (0,1)

getClassificationRateFor2Classes

protected static double getClassificationRateFor2Classes(double[] scoresClass0,
                                                         double[] scoresClass1)

This method computes the classification rate.

Parameters:

scoresClass0 - the scores of class 0 (have to be sorted)

scoresClass1 - the scores of class 1 (have to be sorted)

Returns:

the classification rate