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

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

public class SymmetricalUncertainityBasedFeatureSelection
extends Object

Implementation of the symmetrical uncertanity filter suggested by Lei Yu, Huan Liu in "Feature Selection for High-Dimensional Data: A Fast Correlation-Based Filter Solution" (In "Proceedings of the Twentieth International Conference on Machine Learning", pages 856-863, 2003). This implementation is an adaption of the reference implementation in Weka, with some computational optimizations and without missing value distribution.

Field Summary

private AttributeEvaluationResult[]	attributeEvalResult The results from attribute evaluation
private EvaluationSamples	evalSamples The valuation activites

Constructor Summary

SymmetricalUncertainityBasedFeatureSelection()

Method Summary

AttributeEvaluationResult[]	evaluate(EvaluationSamples evalSamples) Algorithm for FCBF
private AttributeEvaluationResult	evaluateAttribute(int idxAttribute) Evaluates an individual attribute by measuring the symmetrical uncertainty between it and the class.
protected void	prepare(EvaluationSamples evalSamples) Preparation

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

attributeEvalResult

private AttributeEvaluationResult[] attributeEvalResult

The results from attribute evaluation

Constructor Detail

SymmetricalUncertainityBasedFeatureSelection

public SymmetricalUncertainityBasedFeatureSelection()

Method Detail

prepare

protected final void prepare(EvaluationSamples evalSamples)

Preparation

Parameters:

evalSamples - the activity instances with discretized feature values

evaluate

public AttributeEvaluationResult[] evaluate(EvaluationSamples evalSamples)

Algorithm for FCBF

Parameters:

evalSamples - the evaluation training instances with discretized feature values

Returns:

an array (not necessarily ordered) of selected attribute indexes

evaluateAttribute

private AttributeEvaluationResult evaluateAttribute(int idxAttribute)

Evaluates an individual attribute by measuring the symmetrical uncertainty between it and the class.

Parameters:

idxAttribute - the attribute index

Returns:

the evaluation result