javax.datamining.supervised.classification
Interface ReceiverOperatingCharacterics

public interface ReceiverOperatingCharacterics

ReceiverOperatingCharacteristics is a common interface to retrieve the result of receiver operating characteristic computation.

Originally used in signal detection theory to gauge the true hit versus false alarm ratio when sending signals over a noisy channel, the receiver operating characteristics (ROC) of a classifier measures its performance as a trade-off between selectivity and sensitivity. Like lift charts, ROC curves provide a means to compare individual models and determine threshold that result in a high ratio of positive hits.

Typically, an ROC is a curve of false alarm rate (false positives, on X-axis) versus hit rate (true positives, on Y-axis) is plotted with a varying sensitivity or threshold parameter. The curve connects two points (0,0) and (1,1). The point (0,0) is where the classifier always gets the negatives cases right but all positive cases wrong. The point (1,1) is where the classifier gets all positive cases right but all negative cases wrong.

The elements of ROC are the following:

• Probability - Minimum probabilty of the predicted positive class resulting in a positive prediction. Different threshold values result in different hit rates and false alarm rates.
• True positives - Positive cases in the test data with predicted probabilities above the probability threshold (correctly predicted)
• True negatives - Negative cases in the test data with predicted probabilities below the probability threshold (correctly predicted)
• False positives - Negative cases in the test data with predicted probabilities above the probability threshold (incorrectly predicted)
• False negatives - Positive cases in the test data with predicted probabilities below the probability threshold (incorrectly predicted)
• Hit rate - A ratio of true positives to the sum of true positives and false negatives (aka true positive fraction)
• False alarm rate - A ratio of false positives to the sum of false positives and true negatives (aka false positive fraction)

Author:

JSR-73 Java Data Mining Expert Group

Method Summary

double	getAreaUnderCurve() Returns the area under the ROC curve.
double	getFalseAlarmRate(int index) Returns the false alarm rate of the area represented by the specified index.
double	getHitRate(int index) Returns the hit rate of the area represented by the specified index.
long	getNegatives(int index, boolean trueFalse) Returns true/false negatives of the area represented by the specified index.
int	getNumberOfThresholdCandidates() Returns the number of threshold candidates.
long	getPositives(int index, boolean trueFalse) Returns true/false positives of the area represented by the specified index.
double	getProbabilityThreshold(int index) Returns the probability threshold of the area represented by the specified index.

Method Detail

getAreaUnderCurve

public double getAreaUnderCurve()

Returns the area under the ROC curve. The area under the ROC curve is useful as a quantitative measure for the overall performance of models over the entire evaludation dataset. This area ranges between 0 and 1, and an ideal classifier has an area of 1. It represents the probability of the model to correctly rank a randomly chosen pair of rows of opposite classes. If the user wants to use a subset of the evaluation data, the ROC curves also helps in determining a model that will provide the best results at a specific threshold.

Returns:

double

getFalseAlarmRate

public double getFalseAlarmRate(int index)
                         throws JDMException

Returns the false alarm rate of the area represented by the specified index. The index must be between 0 and n-1 where n is the number of threshold candidates.

Parameters:

index -

Returns:

double

Throws:

JDMException

getHitRate

public double getHitRate(int index)
                  throws JDMException

Returns the hit rate of the area represented by the specified index. The index must be between 0 and n-1 where n is the number of threshold candidates.

Parameters:

index -

Returns:

double

Throws:

JDMException

getNegatives

public long getNegatives(int index,
                         boolean trueFalse)
                  throws JDMException

Returns true/false negatives of the area represented by the specified index. The index must be between 0 and n-1 where n is the number of threshold candidates.

Parameters:

index -

trueFalse -

Returns:

long

Throws:

JDMException

getNumberOfThresholdCandidates

public int getNumberOfThresholdCandidates()

Returns the number of threshold candidates. This number can be used as index for other methods.

Returns:

int

getPositives

public long getPositives(int index,
                         boolean trueFalse)
                  throws JDMException

Returns true/false positives of the area represented by the specified index. The index must be between 0 and n-1 where n is the number of threshold candidates.

Parameters:

index -

trueFalse -

Returns:

long

Throws:

JDMException

getProbabilityThreshold

public double getProbabilityThreshold(int index)
                               throws JDMException

Returns the probability threshold of the area represented by the specified index. The index must be between 0 and n-1 where n is the number of threshold candidates.

Parameters:

index -

Returns:

double

Throws:

JDMException