meka.classifiers.multitarget.CCp.java Source code

Introduction

Here is the source code for meka.classifiers.multitarget.CCp.java

Source

/*
 *   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 <http://www.gnu.org/licenses/>.
 */

package meka.classifiers.multitarget;

import java.util.Arrays;
import java.util.Random;

import weka.classifiers.AbstractClassifier;
import meka.classifiers.multilabel.ProblemTransformationMethod;
import weka.core.Instance;
import weka.core.Instances;
import meka.core.MLUtils;
import weka.core.RevisionUtils;
import weka.core.Utils;

/**
 * CCp.java - Multitarget CC with probabilistic output.
 * <br>
 * This version includes probabilistic output in the distributionForInstance, like other MT methods.
 * <br>
 * i.e.: y[j+L] := P(y[j]|x) (this is usefull when used in an ensemble).
 * <br>
 * @see      CC
 * @version   March 2012
 * @author    Jesse Read (jesse@tsc.uc3m.es)
 */

public class CCp extends meka.classifiers.multilabel.CC implements MultiTargetClassifier {

    /** for serialization. */
    private static final long serialVersionUID = 7139310187485439658L;

    protected meka.classifiers.multitarget.CCp.Link root = null;

    protected class Link {

        private meka.classifiers.multitarget.CCp.Link next = null;
        private AbstractClassifier classifier = null;
        public Instances _template = null;
        private int index = -1;
        private int excld[]; // to contain the indices to delete
        private int j = 0; //@temp

        public Link(int chain[], int j, Instances train) throws Exception {
            this.j = j;

            this.index = chain[j];

            // sort out excludes [4|5,1,0,2,3]
            this.excld = Arrays.copyOfRange(chain, j + 1, chain.length);
            // sort out excludes [0,1,2,3,5]
            Arrays.sort(this.excld);

            this.classifier = (AbstractClassifier) AbstractClassifier.forName(getClassifier().getClass().getName(),
                    ((AbstractClassifier) getClassifier()).getOptions());

            Instances new_train = new Instances(train);

            // delete all except one (leaving a binary problem)
            if (getDebug())
                System.out.print(" " + this.index);
            new_train.setClassIndex(-1);
            // delete all the attributes (and track where our index ends up)
            int c_index = chain[j];
            for (int i = excld.length - 1; i >= 0; i--) {
                new_train.deleteAttributeAt(excld[i]);
                if (excld[i] < this.index)
                    c_index--;
            }
            new_train.setClassIndex(c_index);

            _template = new Instances(new_train, 0);

            this.classifier.buildClassifier(new_train);
            new_train = null;

            if (j + 1 < chain.length)
                next = new meka.classifiers.multitarget.CCp.Link(chain, ++j, train);
        }

        protected void classify(Instance test) throws Exception {
            // copy
            Instance copy = (Instance) test.copy();
            copy.setDataset(null);

            // delete attributes we don't need
            for (int i = excld.length - 1; i >= 0; i--) {
                copy.deleteAttributeAt(this.excld[i]);
            }

            //set template
            copy.setDataset(this._template);

            // round
            for (int k = 0; k < this.j; k++) {
                copy.setValue(j, Math.round(copy.value(k)));
            }

            //set class
            double dist[] = this.classifier.distributionForInstance(copy);
            int max_index = Utils.maxIndex(dist);
            confidences[this.index] = dist[max_index];
            test.setValue(this.index, max_index);

            //carry on
            if (next != null)
                next.classify(test);
        }

        @Override
        public String toString() {
            return (next == null) ? String.valueOf(this.index) : String.valueOf(this.index) + ">" + next.toString();
        }

    }

    /**
     * Description to display in the GUI.
     * 
     * @return      the description
     */
    @Override
    public String globalInfo() {
        return "CC method with probabilistic output (CCp).\n"
                + "This version includes probabilistic output in the distributionForInstance, like other MT methods.\n"
                + "i.e.: y[j+L] := P(y[j]|x) (this is usefull when used in an ensemble).";
    }

    @Override
    public void buildClassifier(Instances D) throws Exception {
        testCapabilities(D);

        int L = D.classIndex();

        prepareChain(L);

        if (getDebug())
            System.out.print(":- Chain (");
        root = new meka.classifiers.multitarget.CCp.Link(retrieveChain(), 0, D);
        if (getDebug())
            System.out.println(" ) -:");
    }

    protected double confidences[] = null;

    @Override
    public double[] distributionForInstance(Instance x) throws Exception {
        int L = x.classIndex();
        confidences = new double[L];
        root.classify(x);
        double y[] = new double[L * 2];
        for (int j = 0; j < L; j++) {
            y[j] = x.value(j);
            y[j + L] = confidences[j]; // <--- this is the extra line
        }
        return y;
    }

    @Override
    public String getRevision() {
        return RevisionUtils.extract("$Revision: 9117 $");
    }

    public static void main(String args[]) {
        ProblemTransformationMethod.evaluation(new CCp(), args);
    }
}