cc.mallet.types
Class ROCData

java.lang.Object
  cc.mallet.types.ROCData

All Implemented Interfaces:

AlphabetCarrying, java.io.Serializable

public class ROCData
extends java.lang.Object
implements AlphabetCarrying, java.io.Serializable

Tracks ROC data for instances in Trial results.

Author:

Michael Bond mikejbond@gmail.com

See Also:

Trial, InstanceList, Classifier, Classification, Serialized Form

Field Summary

static int	FALSE_NEGATIVE
static int	FALSE_POSITIVE
static int	TRUE_NEGATIVE
static int	TRUE_POSITIVE

Constructor Summary

ROCData(double[] thresholds, LabelAlphabet labelAlphabet)
Constructs a new object

Method Summary

void	add(Classification classification) Adds classification results to the ROC data
void	add(ROCData rocData) Adds existing ROC data to this ROC data
void	add(Trial trial) Adds trial results to the ROC data
Alphabet	getAlphabet()
Alphabet[]	getAlphabets()
int[][]	getCounts(Label label) Gets the raw counts for a specified label.
int[]	getCounts(Label label, double threshold) Gets the raw counts for a specified label and threshold.
LabelAlphabet	getLabelAlphabet() Gets the label alphabet
double	getPositivePercent(Label label, double threshold) Gets the estimated percentage of training events that exceed the threshold.
double	getPrecision(Label label, double threshold) Gets the precision for a specified label and threshold.
double	getPrecisionForScore(Label label, double score) Gets the precision for a specified label and score.
double	getRecall(Label label, double threshold) Gets the recall rate for a specified label and threshold.
double[]	getThresholds() Gets the thresholds being tracked
void	setCounts(Label label, double threshold, int[] newCounts) Sets the raw counts for a specified label and threshold.
java.lang.String	toString()

Methods inherited from class java.lang.Object

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

Field Detail

TRUE_POSITIVE

public static final int TRUE_POSITIVE

See Also:

Constant Field Values

FALSE_POSITIVE

public static final int FALSE_POSITIVE

See Also:

Constant Field Values

FALSE_NEGATIVE

public static final int FALSE_NEGATIVE

See Also:

Constant Field Values

TRUE_NEGATIVE

public static final int TRUE_NEGATIVE

See Also:

Constant Field Values

Constructor Detail

ROCData

public ROCData(double[] thresholds,
               LabelAlphabet labelAlphabet)

Constructs a new object

Parameters:

thresholds - Array of thresholds to track counts for

labelAlphabet - Label alphabet for instances in Trial

Method Detail

add

public void add(Classification classification)

Adds classification results to the ROC data

Parameters:

trial - Trial results to add to ROC data

add

public void add(Trial trial)

Adds trial results to the ROC data

Parameters:

trial - Trial results to add to ROC data

add

public void add(ROCData rocData)

Adds existing ROC data to this ROC data

Parameters:

rocData - ROC data to add

getAlphabet

public Alphabet getAlphabet()

Specified by:

getAlphabet in interface AlphabetCarrying

getAlphabets

public Alphabet[] getAlphabets()

Specified by:

getAlphabets in interface AlphabetCarrying

getCounts

public int[][] getCounts(Label label)

Gets the raw counts for a specified label.

Parameters:

label - Label to get counts for

Returns:

Array of raw counts for specified label

See Also:

TRUE_POSITIVE, FALSE_POSITIVE, FALSE_NEGATIVE, TRUE_NEGATIVE

getCounts

public int[] getCounts(Label label,
                       double threshold)

Gets the raw counts for a specified label and threshold. If data was not collected for the exact threshold specified, then results for the highest threshold <= the specified threshold will be returned.

Parameters:

label - Label to get counts for

threshold - Threshold to get counts for

Returns:

Array of raw counts for specified label and threshold

See Also:

TRUE_POSITIVE, FALSE_POSITIVE, FALSE_NEGATIVE, TRUE_NEGATIVE

getLabelAlphabet

public LabelAlphabet getLabelAlphabet()

Gets the label alphabet

getPrecision

public double getPrecision(Label label,
                           double threshold)

Gets the precision for a specified label and threshold. If data was not collected for the exact threshold specified, then results will for the highest threshold <= the specified threshold will be returned.

Parameters:

label - Label to get precision for

threshold - Threshold to get precision for

Returns:

Precision for specified label and threshold

getPrecisionForScore

public double getPrecisionForScore(Label label,
                                   double score)

Gets the precision for a specified label and score. This differs from ROCData.getPrecision(Label, double) in that it is the precision for only scores falling in the one score value, not for all scores above the threshold. If data was not collected for the exact threshold specified, then results will for the highest threshold <= the specified threshold will be returned.

Parameters:

label - Label to get precision for

threshold - Threshold to get precision for

Returns:

Precision for specified label and score

getPositivePercent

public double getPositivePercent(Label label,
                                 double threshold)

Gets the estimated percentage of training events that exceed the threshold.

Parameters:

label - Label to get precision for

threshold - Threshold to get precision for

Returns:

Estimated percentage of events exceeding threshold

getRecall

public double getRecall(Label label,
                        double threshold)

Gets the recall rate for a specified label and threshold. If data was not collected for the exact threshold specified, then results will for the highest threshold <= the specified threshold will be returned.

Parameters:

label - Label to get recall for

threshold - Threshold to get recall for

Returns:

Recall rate for specified label and threshold

getThresholds

public double[] getThresholds()

Gets the thresholds being tracked

Returns:

Array of thresholds

setCounts

public void setCounts(Label label,
                      double threshold,
                      int[] newCounts)

Sets the raw counts for a specified label and threshold. If data is not collected for the exact threshold specified, then counts for the highest threshold <= the specified threshold will be set.

Parameters:

label - Label to get counts for

threshold - Threshold to get counts for

newCounts - New count values for the label and threshold

See Also:

TRUE_POSITIVE, FALSE_POSITIVE, FALSE_NEGATIVE, TRUE_NEGATIVE

toString

public java.lang.String toString()

Overrides:

toString in class java.lang.Object