Java Machine Learning Library Source code

/**
 * %SVN.HEADER%
 */
package junit.featureselection;

import net.sf.javaml.core.Dataset;
import net.sf.javaml.core.DefaultDataset;
import net.sf.javaml.core.DenseInstance;
import net.sf.javaml.core.Instance;
import net.sf.javaml.featureselection.FeatureRanking;
import net.sf.javaml.featureselection.ensemble.LinearRankingEnsemble;
import net.sf.javaml.featureselection.ranking.RecursiveFeatureEliminationSVM;
import net.sf.javaml.sampling.Sampling;

import org.junit.Test;

import be.abeel.util.Pair;
/**
 * Test-case for Linear ensemble feature selection
 * 
 * @author Thomas Abeel
 *
 */
public class TestLinearEnsemble {
    @Test
    public void testGreedyBackwardEliminationSynthetic() {

        Dataset data = new DefaultDataset();
        for (int i = 0; i < 100; i++) {
            double[] vals = { i / 50,vary(1, 0), Math.random(), vary(5 * (i / 50), 1) };
            Instance inst = new DenseInstance(vals, ""+(i / 50));
            data.add(inst);

        }
        
        
        
    	FeatureRanking[] ars = new FeatureRanking[10];
		for (int i = 0; i < ars.length; i++)
			ars[i] = new RecursiveFeatureEliminationSVM(1.0);

		Pair<Dataset,Dataset>split=Sampling.SubSampling.sample(data,(int)(data.size()*0.5));
		System.out.println("Training feature selection...");
		FeatureRanking ens = new LinearRankingEnsemble(ars);
		ens.build(split.x().copy());

		for(int i=0;i<data.noAttributes();i++)
			System.out.println(i+"\t"+ens.rank(i));
//        System.out.println(ga.selectedAttributes());
//        Assert.assertTrue(ga.selectedAttributes().contains(0));
//        Assert.assertTrue(ga.selectedAttributes().contains(3));
       

    }

   

    private double vary(double i, double j) {
        return i + j * Math.random();
    }
}