weka.classifiers.evaluation

Interface InformationRetrievalEvaluationMetric

public interface InformationRetrievalEvaluationMetric

An interface for information retrieval evaluation metrics to implement. Allows the command line interface to display these metrics or not based on user-supplied options. These statistics will be displayed as new columns in the table of information retrieval statistics. As such, a toSummaryString() formatted representation is not required.

Version:

$Revision: 9320 $

Author:

Mark Hall (mhall{[at]}pentaho{[dot]}com)

Method Summary

Methods

Modifier and Type	Method and Description
double	getClassWeightedAverageStatistic(String statName) Get the weighted (by class) average for this statistic.
double	getStatistic(String statName, int classIndex) Get the value of the named statistic for the given class index.
void	updateStatsForClassifier(double[] predictedDistribution, Instance instance) Updates the statistics about a classifiers performance for the current test instance.

Method Detail

updateStatsForClassifier

void updateStatsForClassifier(double[] predictedDistribution,
                            Instance instance)
                              throws Exception

Updates the statistics about a classifiers performance for the current test instance. Implementers need only implement this method if it is not possible to compute their statistics from what is stored in the base Evaluation object.

Parameters:

predictedDistribution - the probabilities assigned to each class

instance - the instance to be classified

Throws:

Exception - if the class of the instance is not set

getStatistic

double getStatistic(String statName,
                  int classIndex)

Get the value of the named statistic for the given class index. If the implementing class is extending AbstractEvaluationMetric then the implementation of getStatistic(String statName) should just call this method with a classIndex of 0.

Parameters:

statName - the name of the statistic to compute the value for

classIndex - the class index for which to compute the statistic

Returns:

the value of the named statistic for the given class index or Utils.missingValue() if the statistic can't be computed for some reason

getClassWeightedAverageStatistic

double getClassWeightedAverageStatistic(String statName)

Get the weighted (by class) average for this statistic.

Parameters:

statName - the name of the statistic to compute

Returns:

the weighted (by class) average value of the statistic or Utils.missingValue() if this can't be computed (or isn't appropriate).