weka.classifiers.meta.BatchPredictorVote Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of distributedWekaBase Show documentation
This package provides generic configuration class and distributed map/reduce style tasks for Weka
There is a newer version: 1.0.17
/*
 *   This program is free software: you can redistribute it and/or modify
 *   it under the terms of the GNU General Public License as published by
 *   the Free Software Foundation, either version 3 of the License, or
 *   (at your option) any later version.
 *
 *   This program is distributed in the hope that it will be useful,
 *   but WITHOUT ANY WARRANTY; without even the implied warranty of
 *   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *   GNU General Public License for more details.
 *
 *   You should have received a copy of the GNU General Public License
 *   along with this program.  If not, see .
 */

/*
 *    BatchPredictorVote.java
 *    Copyright (C) 2014 University of Waikato, Hamilton, New Zealand
 *
 */

package weka.classifiers.meta;

import weka.core.BatchPredictor;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Utils;
import weka.gui.GPCIgnore;
import weka.gui.beans.KFIgnore;

/**
 * Class that extends Vote in order to implement BatchPredictor. This is of
 * particular use for forming ensembles of R classifiers learned via the
 * MLRClassifier in the RPlugin.
 * 
 * @author Mark Hall (mhall{[at]}pentaho{[dot]}com)
 * @version $Revision: $
 */
@KFIgnore
@GPCIgnore
public class BatchPredictorVote extends Vote implements BatchPredictor {

  /** For serialization */
  private static final long serialVersionUID = 3807444052796114383L;

  /** Batch prediction size */
  protected String m_batchPredictionSize = "1000";

  @Override
  public void setBatchSize(String size) {
    m_batchPredictionSize = size;
  }

  @Override
  public String getBatchSize() {
    return m_batchPredictionSize;
  }

  @Override
  public double[][] distributionsForInstances(Instances insts) throws Exception {
    // For speed we won't do any checks that the base classifiers are all
    // BatchPredictors

    // TODO if this gets used anywhere but in distributed Weka then the other
    // combination strategies will need to be implemented!!

    return distributionsForInstancesAverage(insts);
  }

  /**
   * Computes the average of probabilities combination rule
   * 
   * @param insts the instances to predict
   * @return an array of averaged predictions
   * @throws Exception if a problem occurs
   */
  protected double[][] distributionsForInstancesAverage(Instances insts)
    throws Exception {
    double[][] preds = new double[insts.numInstances()][insts.numClasses()];
    double[] numPredictions = new double[insts.numInstances()];

    for (int i = 0; i < m_Classifiers.length; i++) {
      double[][] dists =
        ((BatchPredictor) getClassifier(i)).distributionsForInstances(insts);

      for (int j = 0; j < insts.numInstances(); j++) {
        Instance instance = insts.instance(j);
        if (!instance.classAttribute().isNumeric()
          || !Utils.isMissingValue(dists[j][0])) {
          for (int k = 0; k < dists[j].length; k++) {
            preds[j][k] += dists[j][k];
          }
          numPredictions[j]++;
        }
      }
    }

    for (int i = 0; i < m_preBuiltClassifiers.size(); i++) {
      double[][] dists =
        ((BatchPredictor) m_preBuiltClassifiers.get(i))
          .distributionsForInstances(insts);

      for (int j = 0; j < insts.numInstances(); j++) {
        Instance instance = insts.instance(j);
        if (!instance.classAttribute().isNumeric()
          || !Utils.isMissingValue(dists[j][0])) {
          for (int k = 0; k < dists[j].length; k++) {
            preds[j][k] += dists[j][k];
          }
          numPredictions[j]++;
        }
      }
    }

    for (int i = 0; i < insts.numInstances(); i++) {
      if (insts.classAttribute().isNumeric()) {
        if (numPredictions[i] == 0) {
          preds[i][0] = Utils.missingValue();
        } else {
          for (int j = 0; j < preds[i].length; j++) {
            preds[i][j] /= numPredictions[i];
          }
        }
      } else {

        // Should normalize "probability" distribution
        if (Utils.sum(preds[i]) > 0) {
          Utils.normalize(preds[i]);
        }
      }
    }

    return preds;
  }

  @Override
  public boolean implementsMoreEfficientBatchPrediction() {
    for (int i = 0; i < m_Classifiers.length; i++) {
      if (!((BatchPredictor) m_Classifiers[i])
        .implementsMoreEfficientBatchPrediction()) {
        return false;
      }
    }
    return true;
  }
}