org.android.activityminer.logic.features.selection
Class FeatureDiscretizer

java.lang.Object
  org.android.activityminer.logic.features.selection.FeatureDiscretizer

public class FeatureDiscretizer
extends Object

Implementation of a discretizer for continuous real attributes following the Fayyad & Irani MDL method. The implementation is an adaption from Weka for the android platform (http://grepcode.com/file/repo1 .maven.org/maven2/nz.ac.waikato.cms.weka/weka-dev /3.7.10/weka/filters/supervised/attribute/Discretize.java#Discretize).

Field Summary

private ActivityByAttributeComparator	activityComparator The attribute value comparator
private double[][]	cutPoints Cut-points
private EvaluationSamples	evalSamples The valuation activites

Constructor Summary

FeatureDiscretizer()

Method Summary

protected void	calculateCutPoints() Cut point calculation
private double[]	calculateCutPointsForAttribute(Activity[] instances, int idxAttribute, int idxStart, int idxStop, AttributeValueComparator attributeComparator) Does compute the cut points for a given attribute and a given activity index range
private void	convert(Activity activity) Does convert a activity
EvaluationSamples	convert(Activity[] activities) Does convert the features of the given activities to categorical attribute values
double[][]	getCutPoints() Getter for the cutPoints
protected void	prepare(Activity[] activities) Preparation with training instances
private boolean	testWithFayyadAndIranisMDL(double[] priorCounts, double priorEntropy, double[][] bestCounts, int cntActivities, int cntCutPoints) Test with Fayyad and Irani's MDL criterion

Methods inherited from class java.lang.Object

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

Field Detail

evalSamples

private EvaluationSamples evalSamples

The valuation activites

cutPoints

private double[][] cutPoints

Cut-points

activityComparator

private final ActivityByAttributeComparator activityComparator

The attribute value comparator

Constructor Detail

FeatureDiscretizer

public FeatureDiscretizer()

Method Detail

prepare

protected final void prepare(Activity[] activities)

Preparation with training instances

Parameters:

activities - the activity instances

convert

public final EvaluationSamples convert(Activity[] activities)

Does convert the features of the given activities to categorical attribute values

Parameters:

activities - the activities for feature conversion

Returns:

the array with converted activities sorted by class identifiers in ascending order

convert

private final void convert(Activity activity)

Does convert a activity

Parameters:

activity - the activity to convert

calculateCutPoints

protected final void calculateCutPoints()

Cut point calculation

calculateCutPointsForAttribute

private double[] calculateCutPointsForAttribute(Activity[] instances,
                                                int idxAttribute,
                                                int idxStart,
                                                int idxStop,
                                                AttributeValueComparator attributeComparator)

Does compute the cut points for a given attribute and a given activity index range

Parameters:

instances - the activities sorted by attribute index

idxAttribute - the attribute index

idxStart - the index to start from

idxStop - the index behind the last activity to use for cut-point calculation

attributeComparator - the attribute comparator

Returns:

the cut points for the attribute index

testWithFayyadAndIranisMDL

private final boolean testWithFayyadAndIranisMDL(double[] priorCounts,
                                                 double priorEntropy,
                                                 double[][] bestCounts,
                                                 int cntActivities,
                                                 int cntCutPoints)

Test with Fayyad and Irani's MDL criterion

Parameters:

priorCounts - the prior counts

priorEntropy - the prior entropy

bestCounts - the best counts

cntActivities - the activity count

cntCutPoints - the current cut point count

Returns:

true if split is accepted, false otherwise

getCutPoints

public final double[][] getCutPoints()

Getter for the cutPoints

Returns:

the cutPoints