org.android.activityminer.logic.evaluation
Class EvaluationHelper

java.lang.Object
  de.unikassel.android.sdcframework.util.ObservableEventSourceImpl<ObservableProgressData>
      org.android.activityminer.logic.evaluation.EvaluationHelper

All Implemented Interfaces:

ObservableEventSource<ObservableProgressData>, ObserverRegistration<ObservableProgressData>

public class EvaluationHelper
extends ObservableEventSourceImpl<ObservableProgressData>

A class to delegate evaluation processing to.

Field Summary

private boolean	bestResultsOnly The flag if only the best results shall be returned.
private Configuration	config The current configuration.
private Context	context The context.
private DBManager	dbManager A reference to the database manager.
private EvaluationResultCollection	evaluationResults The evaluation results.
private List<Set<FeatureTypes>>	featureSetsToTest The feature sets to test.
private String	lastError The last error
private android.support.v4.util.LongSparseArray<List<Long>>	mapSelectedClassId2SampleIds A map to store sample (feature vector) identifiers for the selected lass identifiers
private double	maxPredictionTime The maximum prediction time for a sample.
private double	minPredictionTime The minimum prediction time for a sample.
private int	progress The current progress.
private ObservableProgressData	progressData The observable progress data
private int	sampleCount The total sample count.
private boolean	succeeded The success state flag
private double	sumOfAvgPredictionTime The sum of average prediction time for a sample.
private double	sumOfAvgTrainingTime The sum of average training time for a sample.
private double	threshold The accuracy threshold in percent (0%-100%).
private List<TrainingActivity>	trainingActivites The training activities for evaluation.
private ValidationStrategy	validationStrategy The validation strategy.

Constructor Summary

EvaluationHelper(DBManager dbManager, Set<Set<FeatureTypes>> featureSetsToTest, Configuration config, boolean bestResultsOnly, Context context)
Constructor

Method Summary

long	doEvaluate() Does run the evaluation for all feature sets
void	evaluate(Set<FeatureTypes> selectedFeatures, Classifier classifier) Evaluation for a single feature set.
EvaluationResultCollection	getEvaluationResults() Getter for the evaluation results
String	getLastError() Getter for the last error
int	getMaxProgress() Getter for the maximum progress
private static List<Set<FeatureTypes>>	getSetsSortedBySize(Set<Set<FeatureTypes>> featureSets) Does sort the feature sets for evaluation by size.
String	getStartProgressMessage() Getter for the start progress message
double	getThreshold() Getter for the accuracy threshold
boolean	hasSucceeded() Getter for the success flag
boolean	isBestResultsOnly() Getter for the flag for selection of best results
void	loadSamplesAndCreatePartioning() Does load available samples as feature vectors from DB and creates the partitioning for k-fold cross validation.
boolean	saveResultsToFile(String fileName) Method to store evaluation result to file.
void	updateProgress(String msg) Does publish the current progress with the given message.

Methods inherited from class de.unikassel.android.sdcframework.util.ObservableEventSourceImpl

getObservers, hasObservers, notify, onObserverRegistration, onObserverUnregistration, registerEventObserver, removeAllObservers, unregisterEventObserver

Methods inherited from class java.lang.Object

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

Field Detail

dbManager

private final DBManager dbManager

A reference to the database manager.

config

private Configuration config

The current configuration.

featureSetsToTest

private final List<Set<FeatureTypes>> featureSetsToTest

The feature sets to test.

sampleCount

private int sampleCount

The total sample count.

validationStrategy

private final ValidationStrategy validationStrategy

The validation strategy.

trainingActivites

private List<TrainingActivity> trainingActivites

The training activities for evaluation.

evaluationResults

private final EvaluationResultCollection evaluationResults

The evaluation results.

sumOfAvgTrainingTime

private double sumOfAvgTrainingTime

The sum of average training time for a sample.

sumOfAvgPredictionTime

private double sumOfAvgPredictionTime

The sum of average prediction time for a sample.

minPredictionTime

private double minPredictionTime

The minimum prediction time for a sample.

maxPredictionTime

private double maxPredictionTime

The maximum prediction time for a sample.

progressData

private final ObservableProgressData progressData

The observable progress data

progress

private int progress

The current progress.

succeeded

private boolean succeeded

The success state flag

lastError

private String lastError

The last error

mapSelectedClassId2SampleIds

private final android.support.v4.util.LongSparseArray<List<Long>> mapSelectedClassId2SampleIds

A map to store sample (feature vector) identifiers for the selected lass identifiers

context

private Context context

The context.

bestResultsOnly

private final boolean bestResultsOnly

The flag if only the best results shall be returned.

threshold

private double threshold

The accuracy threshold in percent (0%-100%). If bestResultsOnly is false, this parameter limits the results by a minimal expected overall accuracy value

Constructor Detail

EvaluationHelper

public EvaluationHelper(DBManager dbManager,
                        Set<Set<FeatureTypes>> featureSetsToTest,
                        Configuration config,
                        boolean bestResultsOnly,
                        Context context)

Constructor

Parameters:

dbManager - a reference to the database manager

featureSetsToTest - the set with the feature sets to test

config - the current configuration

bestResultsOnly - flag to limit the result list to the best results (ranked by overall prediction accuracy)

context - the context

Method Detail

getThreshold

public double getThreshold()

Getter for the accuracy threshold

Returns:

the accuracy threshold

isBestResultsOnly

public boolean isBestResultsOnly()

Getter for the flag for selection of best results

Returns:

true if only the best results have been selected, false otherwise

hasSucceeded

public final boolean hasSucceeded()

Getter for the success flag

Returns:

the success flag

getLastError

public final String getLastError()

Getter for the last error

Returns:

the last error

getSetsSortedBySize

private static final List<Set<FeatureTypes>> getSetsSortedBySize(Set<Set<FeatureTypes>> featureSets)

Does sort the feature sets for evaluation by size.

Parameters:

featureSets - the set with the feature sets

Returns:

a list with the feature sets sorted by size in ascending order

getMaxProgress

public final int getMaxProgress()

Getter for the maximum progress

Returns:

the maximum progress steps

getEvaluationResults

public final EvaluationResultCollection getEvaluationResults()

Getter for the evaluation results

Returns:

the evaluation results

doEvaluate

public final long doEvaluate()

Does run the evaluation for all feature sets

Returns:

the count of successfully evaluated feature sets

saveResultsToFile

public final boolean saveResultsToFile(String fileName)

Method to store evaluation result to file.

Parameters:

fileName - the filename

Returns:

true if successful, false otherwise

loadSamplesAndCreatePartioning

public final void loadSamplesAndCreatePartioning()
                                          throws Exception

Does load available samples as feature vectors from DB and creates the partitioning for k-fold cross validation.

Throws:

Exception

evaluate

public final void evaluate(Set<FeatureTypes> selectedFeatures,
                           Classifier classifier)

Evaluation for a single feature set.

Parameters:

selectedFeatures - the selected features for evaluation

classifier - the classifier for evaluation

updateProgress

public final void updateProgress(String msg)

Does publish the current progress with the given message.

Parameters:

msg - the progress message

getStartProgressMessage

public String getStartProgressMessage()

Getter for the start progress message

Returns:

the start message for progress observation