source: trunk/CrossPare/src/de/ugoe/cs/cpdp/dataprocessing/Undersampling.java @ 23

package de.ugoe.cs.cpdp.dataprocessing;

import org.apache.commons.collections4.list.SetUniqueList;

import weka.core.Instances;
import weka.filters.Filter;
import weka.filters.supervised.instance.Resample;

/**
 * Implements undersampling, a strategy for handling bias in data. In case there are less positive samples (i.e. defect-prone) samples in the
 * data than negative samples (i.e. non-defect-prone), the non-defect-prone entities are sampled such thatthe number of defect-prone and non-defect-prone instances is the same afterwards.  
 * @author Steffen Herbold
 */
public class Undersampling implements IProcessesingStrategy,
                ISetWiseProcessingStrategy {

        
        /**
         * Does not have parameters. String is ignored.
         * @param parameters ignored
         */
        @Override
        public void setParameter(String parameters) {
                // dummy
        }

        /*
         * (non-Javadoc)
         * @see de.ugoe.cs.cpdp.dataprocessing.ISetWiseProcessingStrategy#apply(weka.core.Instances, org.apache.commons.collections4.list.SetUniqueList)
         */
        @Override
        public void apply(Instances testdata, SetUniqueList<Instances> traindataSet) {
                for( Instances traindata : traindataSet ) {
                        apply(testdata, traindata);
                }
        }

        /*
         * (non-Javadoc)
         * @see de.ugoe.cs.cpdp.dataprocessing.IProcessesingStrategy#apply(weka.core.Instances, weka.core.Instances)
         */
        @Override
        public void apply(Instances testdata, Instances traindata) {
                
                final int[] counts = traindata.attributeStats(traindata.classIndex()).nominalCounts;
                
                if( counts[1]<counts[0] ) {
                        Instances negatives = new Instances(traindata);
                        Instances positives = new Instances(traindata);
                        
                        for( int i=traindata.size()-1 ; i>=0 ; i-- ) {
                                if( Double.compare(1.0, negatives.get(i).classValue())==0 ) {
                                        negatives.remove(i);
                                }
                                if( Double.compare(0.0, positives.get(i).classValue())==0 ) {
                                        positives.remove(i);
                                }
                        }
                        
                        Resample resample = new Resample();
                        // TODO: resample.setSampleSizePercent((100.0*counts[1])/100+0.01);
                        // Ohne +0.01 wird bei tomcat, xerces-1.2 und jedit-4.0 ein negative weniger zurückgegeben
                        resample.setSampleSizePercent((100.0* counts[1])/counts[0]);
                        try {
                                resample.setInputFormat(traindata);
                                negatives = Filter.useFilter(negatives, resample);
                        } catch (Exception e) {
                                throw new RuntimeException(e);
                        }
                        traindata.clear();
                        for( int i=0 ; i<negatives.size() ; i++ ) {
                                traindata.add(negatives.get(i));
                        }
                        for( int i=0 ; i<positives.size() ; i++ ) {
                                traindata.add(positives.get(i));
                        }
                }
        }

}