adams.data.weka.evaluator
Class IntervalEstimatorBased

java.lang.Object
  adams.core.ConsoleObject
      adams.core.option.AbstractOptionHandler
          adams.data.weka.evaluator.AbstractInstanceEvaluator
              adams.data.weka.evaluator.AbstractDatasetInstanceEvaluator
                  adams.data.weka.evaluator.AbstractCrossvalidatedInstanceEvaluator<IntervalEstimatorBased.SortedInterval>
                      adams.data.weka.evaluator.IntervalEstimatorBased

All Implemented Interfaces:

CleanUpHandler, Debuggable, Destroyable, OptionHandler, SizeOfHandler, Serializable

public class IntervalEstimatorBased
extends AbstractCrossvalidatedInstanceEvaluator<IntervalEstimatorBased.SortedInterval>

Uses a classifier that produces confidence intervals. ???

Valid options are:

-D <int> (property: debugLevel)
    The greater the number the more additional info the scheme may output to
    the console (0 = off).
    default: 0
    minimum: 0
 

-threshold <double> (property: threshold)
    The threshold percentage to use (0-1).
    default: 0.75
 

-folds <int> (property: folds)
    The number of folds to use for cross-validation; cross-validation gets turned
    off below a value of 2.
    default: 2
 

-seed <int> (property: seed)
    The seed value for randomizing the data for cross-validation.
    default: 1
 

-classifier <weka.classifiers.Classifier [options]> (property: classifier)
    The classifier to use (must implement weka.classifiers.IntervalEstimator
    ).
    default: weka.classifiers.functions.GaussianProcesses -L 1.0 -N 0 -K \"weka.classifiers.functions.supportVector.PolyKernel -C 250007 -E 1.0\"
 

-level <double> (property: confidenceLevel)
    The confidence level to use when generating the confidence intervals (0-
    1).
    default: 0.95
 

-relative (property: relativeWidths)
    If set to true, then the calculated widths will be relative ones, as they
    will get divided by the class value of the Instance.
 

Version:

$Revision: 4584 $

Author:

fracpete (fracpete at waikato dot ac dot nz)

See Also:

Serialized Form

Nested Class Summary

static class

IntervalEstimatorBased.SortedInterval Helper class for sorting the confidence intervals.

Nested classes/interfaces inherited from class adams.data.weka.evaluator.AbstractCrossvalidatedInstanceEvaluator

AbstractCrossvalidatedInstanceEvaluator.EvaluationContainer

Field Summary

protected weka.classifiers.Classifier	m_Classifier the IntervalEstimator to use.
protected double	m_ConfidenceLevel the confidence level.
protected double	m_MaxWidth the maximum width allowed.
protected double	m_MinWidth the minimum width encountered.
protected boolean	m_RelativeWidths whether to divide the calculated widths by the class value.

Fields inherited from class adams.data.weka.evaluator.AbstractCrossvalidatedInstanceEvaluator

m_Folds, m_Seed

Fields inherited from class adams.data.weka.evaluator.AbstractDatasetInstanceEvaluator

m_Data, m_Threshold

Fields inherited from class adams.data.weka.evaluator.AbstractInstanceEvaluator

m_Initialized

Fields inherited from class adams.core.option.AbstractOptionHandler

m_DebugLevel, m_OptionManager

Constructor Summary

IntervalEstimatorBased()

Method Summary

protected static double	calcAverageWidth(double[][] array) Calculates the average width of the intervals.
protected static double	calcWidth(double[] array) Calculates the width of the interval.
String	classifierTipText() Returns the tip text for this property.
String	confidenceLevelTipText() Returns the tip text for this property.
void	defineOptions() Adds options to the internal list of options.
protected double	doEvaluate(weka.core.Instance inst) Performs the actual evaluation.
protected Vector<IntervalEstimatorBased.SortedInterval>	evaluate(weka.core.Instances train, weka.core.Instances test) Performs an evaluation on the given train and test set.
protected String	findThreshold(Vector<IntervalEstimatorBased.SortedInterval> evals) Finds the threshold based on the collected data.
weka.classifiers.Classifier	getClassifier() Returns the classifier.
double	getConfidenceLevel() Returns the confidence level.
boolean	getRelativeWidths() Returns whether the calculated widths are divided by the class value.
String	globalInfo() Returns a string describing the object.
String	relativeWidthsTipText() Returns the tip text for this property.
void	setClassifier(weka.classifiers.Classifier value) Sets the classifier to use, must implement weka.classifiers.IntervalEstimator.
void	setConfidenceLevel(double value) Sets the confidence level.
void	setRelativeWidths(boolean value) Sets whether to divide the calculated widths by the class value.

Methods inherited from class adams.data.weka.evaluator.AbstractCrossvalidatedInstanceEvaluator

findThreshold, foldsTipText, getFolds, getSeed, seedTipText, setFolds, setSeed

Methods inherited from class adams.data.weka.evaluator.AbstractDatasetInstanceEvaluator

cleanUp, getData, getThreshold, reset, setData, setThreshold, setUp, split, thresholdTipText

Methods inherited from class adams.data.weka.evaluator.AbstractInstanceEvaluator

check, compareTo, destroy, equals, evaluate, forCommandLine, forName, getEvaluators, shallowCopy, shallowCopy

Methods inherited from class adams.core.option.AbstractOptionHandler

cleanUpOptions, debug, debug, debugLevelTipText, finishInit, getDebugLevel, getOptionManager, initialize, isDebugOn, newOptionManager, setDebugLevel, toCommandLine, toString

Methods inherited from class adams.core.ConsoleObject

getDebugging, getSystemErr, getSystemOut, sizeOf

Methods inherited from class java.lang.Object

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

Field Detail

m_Classifier

protected weka.classifiers.Classifier m_Classifier

the IntervalEstimator to use.

m_ConfidenceLevel

protected double m_ConfidenceLevel

the confidence level.

m_RelativeWidths

protected boolean m_RelativeWidths

whether to divide the calculated widths by the class value.

m_MaxWidth

protected double m_MaxWidth

the maximum width allowed.

m_MinWidth

protected double m_MinWidth

the minimum width encountered.

Constructor Detail

IntervalEstimatorBased

public IntervalEstimatorBased()

Method Detail

globalInfo

public String globalInfo()

Returns a string describing the object.

Specified by:

globalInfo in class AbstractOptionHandler

Returns:

a description suitable for displaying in the gui

defineOptions

public void defineOptions()

Adds options to the internal list of options.

Specified by:

defineOptions in interface OptionHandler

Overrides:

defineOptions in class AbstractCrossvalidatedInstanceEvaluator<IntervalEstimatorBased.SortedInterval>

setClassifier

public void setClassifier(weka.classifiers.Classifier value)

Sets the classifier to use, must implement weka.classifiers.IntervalEstimator.

Parameters:

value - the classifier

getClassifier

public weka.classifiers.Classifier getClassifier()

Returns the classifier.

Returns:

the classifier

classifierTipText

public String classifierTipText()

Returns the tip text for this property.

Returns:

tip text for this property suitable for displaying in the GUI or for listing the options.

setConfidenceLevel

public void setConfidenceLevel(double value)

Sets the confidence level.

Parameters:

value - the confidence level (0-1)

getConfidenceLevel

public double getConfidenceLevel()

Returns the confidence level.

Returns:

the confidence level (0-1)

confidenceLevelTipText

public String confidenceLevelTipText()

Returns the tip text for this property.

Returns:

tip text for this property suitable for displaying in the GUI or for listing the options.

setRelativeWidths

public void setRelativeWidths(boolean value)

Sets whether to divide the calculated widths by the class value.

Parameters:

value - if true then the widths will get divided by the class value (= relative)

getRelativeWidths

public boolean getRelativeWidths()

Returns whether the calculated widths are divided by the class value.

Returns:

trye if the widths are divided by the class value

relativeWidthsTipText

public String relativeWidthsTipText()

Returns the tip text for this property.

Returns:

tip text for this property suitable for displaying in the GUI or for listing the options.

findThreshold

protected String findThreshold(Vector<IntervalEstimatorBased.SortedInterval> evals)

Finds the threshold based on the collected data.

Specified by:

findThreshold in class AbstractCrossvalidatedInstanceEvaluator<IntervalEstimatorBased.SortedInterval>

Parameters:

evals - the collected evaluation containers

Returns:

null if everything OK, error message otherwise

evaluate

protected Vector<IntervalEstimatorBased.SortedInterval> evaluate(weka.core.Instances train,
                                                                 weka.core.Instances test)

Performs an evaluation on the given train and test set.

Specified by:

evaluate in class AbstractCrossvalidatedInstanceEvaluator<IntervalEstimatorBased.SortedInterval>

Parameters:

train - the training set

test - the test set

Returns:

the generated evaluation container

doEvaluate

protected double doEvaluate(weka.core.Instance inst)

Performs the actual evaluation.

Specified by:

doEvaluate in class AbstractInstanceEvaluator

Parameters:

inst - the instance to evaluate

Returns:

evaluation range, between 0 and 1 (0 = bad, 1 = good, -1 = if unable to evaluate)

calcWidth

protected static double calcWidth(double[] array)

Calculates the width of the interval.

Parameters:

array - the lower and upper bound

Returns:

the width

calcAverageWidth

protected static double calcAverageWidth(double[][] array)

Calculates the average width of the intervals.

Parameters:

array - the arrayw with the lower and upper bounds

Returns:

the average width