StationaryBagging.java

package quickml.supervised.tree.bagging;

import com.google.common.collect.Lists;
import quickml.collections.MapUtils;
import quickml.data.instances.InstanceWithAttributesMap;

import java.io.Serializable;
import java.util.HashSet;
import java.util.List;

/**
 * Created by alexanderhawk on 4/5/15.
 */
public class StationaryBagging implements Bagging {

    private static com.twitter.common.util.Random rand = com.twitter.common.util.Random.Util.fromSystemRandom(MapUtils.random);

    @Override
    public <L extends Serializable,E extends InstanceWithAttributesMap<L>> TrainingDataPair<L, E> separateTrainingDataFromOutOfBagData(List<E> trainingData) {
        List<E> baggedTrainingData = Lists.newArrayList();
        List<E> outOfBagTrainingData = Lists.newArrayList();

        HashSet<Integer> unusedDataIndices = new HashSet<>();
        for (int i = 0; i < trainingData.size(); i++) {
            unusedDataIndices.add(i);
        }
        for (int i = 0; i < trainingData.size(); i++) {
            int toAdd = rand.nextInt(trainingData.size());
            if (unusedDataIndices.contains(toAdd))
                unusedDataIndices.remove(toAdd);
            baggedTrainingData.add(trainingData.get(toAdd));
        }
        for (Integer index : unusedDataIndices) {
            outOfBagTrainingData.add(trainingData.get(index));
        }
        return new TrainingDataPair<>(baggedTrainingData, outOfBagTrainingData);

    }
}