Example usage for weka.core Instances resampleWithWeights

Introduction

In this page you can find the example usage for weka.core Instances resampleWithWeights.

Prototype

public Instances resampleWithWeights(Random random) 

Source Link

Document

Creates a new dataset of the same size as this dataset using random sampling with replacement according to the current instance weights.

Usage

From source file:fantail.algorithms.ARTForests.java

License:Open Source License

@Override
public void buildRanker(Instances metaData) throws Exception {

    Random r = new Random(m_RandomSeed);
    Instances workingData = new Instances(metaData);
    m_WeakRankers = new Ranker[m_T];

    for (int i = 0; i < m_T; i++) {
        Instances baggingSample = workingData.resampleWithWeights(r);

        if (m_BaggingPercentage < 100.0) {
            weka.filters.unsupervised.instance.Resample res = new weka.filters.unsupervised.instance.Resample();
            res.setSampleSizePercent(m_BaggingPercentage);
            res.setNoReplacement(false);
            res.setInputFormat(baggingSample);
            baggingSample = Filter.useFilter(baggingSample, res);
        }/*from   w w  w. j a  va 2  s .co m*/

        BinaryART ranker = new BinaryART();
        ranker.setMiniLeaf(m_NumMinInstances);
        ranker.setK(m_K);
        ranker.setRandomSeed(i);
        ranker.setUseMedian(m_UseMedian);

        m_WeakRankers[i] = ranker;
        m_WeakRankers[i].buildRanker(baggingSample);
    }
}

---

From source file:mulan.classifier.transformation.EnsembleOfClassifierChains.java

License:Open Source License

@Override
protected void buildInternal(MultiLabelInstances trainingSet) throws Exception {

    Instances dataSet = new Instances(trainingSet.getDataSet());

    for (int i = 0; i < numOfModels; i++) {
        debug("ECC Building Model:" + (i + 1) + "/" + numOfModels);
        // 2013.12.13  
        System.out.println("ECC Building Model:" + (i + 1) + "/" + numOfModels);
        Instances sampledDataSet = null;
        dataSet.randomize(rand);//from  w  ww.  j  a  v a2  s .c  om
        if (useSamplingWithReplacement) {
            int bagSize = dataSet.numInstances() * BagSizePercent / 100;
            // create the in-bag dataset
            sampledDataSet = dataSet.resampleWithWeights(new Random(1));
            if (bagSize < dataSet.numInstances()) {
                sampledDataSet = new Instances(sampledDataSet, 0, bagSize);
            }
        } else {
            RemovePercentage rmvp = new RemovePercentage();
            rmvp.setInvertSelection(true);
            rmvp.setPercentage(samplingPercentage);
            rmvp.setInputFormat(dataSet);
            sampledDataSet = Filter.useFilter(dataSet, rmvp);
        }
        MultiLabelInstances train = new MultiLabelInstances(sampledDataSet, trainingSet.getLabelsMetaData());

        int[] chain = new int[numLabels];
        for (int j = 0; j < numLabels; j++)
            chain[j] = j;
        for (int j = 0; j < chain.length; j++) {
            int randomPosition = rand.nextInt(chain.length);
            int temp = chain[j];
            chain[j] = chain[randomPosition];
            chain[randomPosition] = temp;
        }
        debug(Arrays.toString(chain));
        //========================================
        System.out.println(Arrays.toString(chain));
        //========================================
        // MAYBE WE SHOULD CHECK NOT TO PRODUCE THE SAME VECTOR FOR THE
        // INDICES
        // BUT IN THE PAPER IT DID NOT MENTION SOMETHING LIKE THAT
        // IT JUST SIMPLY SAY A RANDOM CHAIN ORDERING OF L

        ensemble[i] = new ClassifierChain(baseClassifier, chain);
        ensemble[i].build(train);
    }

}

---