org.openimaj.experiment.evaluation.classification

Class ClassificationEvaluator<RESULT extends AnalysisResult,CLASS,OBJECT>

java.lang.Object

org.openimaj.experiment.evaluation.classification.ClassificationEvaluator<RESULT,CLASS,OBJECT>

Type Parameters:

RESULT - Type of analysed data

CLASS - Type of classes predicted by the classifier

OBJECT - Type of objects classified by the classifier

All Implemented Interfaces:

Evaluator<Map<OBJECT,ClassificationResult<CLASS>>,RESULT>

public class ClassificationEvaluator<RESULT extends AnalysisResult,CLASS,OBJECT>
extends Object
implements Evaluator<Map<OBJECT,ClassificationResult<CLASS>>,RESULT>

Implementation of an Evaluator for the evaluation of classification experiments.

Author:

Jonathon Hare (jsh2@ecs.soton.ac.uk)

Field Summary

Fields

Modifier and Type	Field and Description
protected Map<OBJECT,Set<CLASS>>	actual
protected ClassificationAnalyser<RESULT,CLASS,OBJECT>	analyser
protected Classifier<CLASS,OBJECT>	classifier
protected Collection<OBJECT>	objects

Constructor Summary

Constructors

Constructor and Description
ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier, Collection<OBJECT> objects, Map<OBJECT,Set<CLASS>> actual, ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser) Construct a new ClassificationEvaluator with the given classifier, set of objects to classify, ground truth ("actual") data and an ClassificationAnalyser.
ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier, GroupedDataset<CLASS,? extends ListDataset<OBJECT>,OBJECT> actual, ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser) Construct a new ClassificationEvaluator with the given classifier, ground truth ("actual") data and an ClassificationAnalyser.
ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier, Map<OBJECT,Set<CLASS>> actual, ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser) Construct a new ClassificationEvaluator with the given classifier, ground truth ("actual") data and an ClassificationAnalyser.
ClassificationEvaluator(Map<OBJECT,ClassificationResult<CLASS>> results, Map<OBJECT,Set<CLASS>> actual, ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser) Construct a new ClassificationEvaluator with the given pre-classified results, the ground truth ("actual") data and an ClassificationAnalyser.

Method Summary

Modifier and Type	Method and Description
RESULT	analyse(Map<OBJECT,ClassificationResult<CLASS>> predicted) Analyse the raw data produced by a call to Evaluator.evaluate() and return the analysed data.
Map<OBJECT,ClassificationResult<CLASS>>	evaluate() Run, or evaluate the system, returning the raw data.
Map<OBJECT,Set<CLASS>>	getExpected() Get the expected classes for each instance

Methods inherited from class java.lang.Object

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

Field Detail

classifier

protected Classifier<CLASS,OBJECT> classifier

analyser

protected ClassificationAnalyser<RESULT extends AnalysisResult,CLASS,OBJECT> analyser

actual

protected Map<OBJECT,Set<CLASS>> actual

objects

protected Collection<OBJECT> objects

Constructor Detail

ClassificationEvaluator

public ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier,
                               Collection<OBJECT> objects,
                               Map<OBJECT,Set<CLASS>> actual,
                               ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser)

Construct a new ClassificationEvaluator with the given classifier, set of objects to classify, ground truth ("actual") data and an ClassificationAnalyser.

Parameters:

classifier - the classifier

objects - the objects to classify

actual - the ground truth

analyser - the analyser

ClassificationEvaluator

public ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier,
                               Map<OBJECT,Set<CLASS>> actual,
                               ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser)

Construct a new ClassificationEvaluator with the given classifier, ground truth ("actual") data and an ClassificationAnalyser.

The objects to classify are taken from the Map.keySet() of the ground truth.

Parameters:

classifier - the classifier

actual - the ground truth

analyser - the analyser

ClassificationEvaluator

public ClassificationEvaluator(Classifier<CLASS,OBJECT> classifier,
                               GroupedDataset<CLASS,? extends ListDataset<OBJECT>,OBJECT> actual,
                               ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser)

Construct a new ClassificationEvaluator with the given classifier, ground truth ("actual") data and an ClassificationAnalyser.

The ground-truth classes to are taken from the GroupedDataset.getGroups() of the "actual" GroupedDataset , and the objects are assembled by concatenating all of the ListDatasets within the "actual" dataset.

Parameters:

classifier - the classifier

actual - the dataset containing instances and ground truths

analyser - the analyser

ClassificationEvaluator

public ClassificationEvaluator(Map<OBJECT,ClassificationResult<CLASS>> results,
                               Map<OBJECT,Set<CLASS>> actual,
                               ClassificationAnalyser<RESULT,CLASS,OBJECT> analyser)

Construct a new ClassificationEvaluator with the given pre-classified results, the ground truth ("actual") data and an ClassificationAnalyser.

Internally, this constructor wraps a simple Classifier implementation around the results.

Parameters:

results - the pre-classified results

actual - the ground truth

analyser - the analyser

Method Detail

evaluate

public Map<OBJECT,ClassificationResult<CLASS>> evaluate()

Description copied from interface: Evaluator

Run, or evaluate the system, returning the raw data.

Specified by:

evaluate in interface Evaluator<Map<OBJECT,ClassificationResult<CLASS>>,RESULT extends AnalysisResult>

Returns:

the raw data

analyse

public RESULT analyse(Map<OBJECT,ClassificationResult<CLASS>> predicted)

Description copied from interface: Evaluator

Analyse the raw data produced by a call to Evaluator.evaluate() and return the analysed data.

Specified by:

analyse in interface Evaluator<Map<OBJECT,ClassificationResult<CLASS>>,RESULT extends AnalysisResult>

Parameters:

predicted - the raw data.

Returns:

the analysed data

getExpected

public Map<OBJECT,Set<CLASS>> getExpected()

Get the expected classes for each instance

Returns:

the map of instances to expected classes