moa.classifiers.meta.PairedLearners Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of moa Show documentation
Massive On-line Analysis is an environment for massive data mining. MOA provides a framework for data stream mining and includes tools for evaluation and a collection of machine learning algorithms. Related to the WEKA project, also written in Java, while scaling to more demanding problems.
The newest version!
/*
 *    PairedLearners.java
 *    Copyright (C) 2015 Instituto Federal de Pernambuco, Recife, Brazil
 *    @author Paulo Gonçalves ([email protected])
 *
 *    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 .
 */
package moa.classifiers.meta;

import com.github.javacliparser.FloatOption;
import com.github.javacliparser.IntOption;
import java.util.Arrays;
import java.util.LinkedList;
import java.util.List;

import moa.classifiers.AbstractClassifier;
import moa.classifiers.Classifier;
import moa.classifiers.MultiClassClassifier;
import moa.core.Measurement;
import moa.options.ClassOption;
import com.yahoo.labs.samoa.instances.Instance;
import moa.core.MiscUtils;

/**
 * Creates two classifiers: a stable and a reactive. The first represents the
 * actual stable concept, while the second is trained on the most recent data.
 * If the accuracy of the reactive is higher than that of the stable, it means
 * the concept has changed. The stable classifier is then substituted by the
 * reactive, and the reactive is reset.
 *
 * 
 * Stephen H. Bach, Marcus A. Maloof, "Paired Learners for Concept Drift",
 * Eighth IEEE International Conference on Data Mining (ICDM), 2008,
 * pp.23-32
 *
 * @author Paulo Gonçalves (paulogoncalves at recife.ifpe.edu.br)
 *
 */

public class PairedLearners extends AbstractClassifier implements MultiClassClassifier {
    private static final long serialVersionUID = 1L;

    public ClassOption stableLearnerOption = new ClassOption("stableLearner", 
            's', "Stable learner", Classifier.class, "bayes.NaiveBayes");

    public ClassOption reactiveLearnerOption = new ClassOption("reactiveLearner", 
            'r', "Reactive learner", Classifier.class, "bayes.NaiveBayes");

    public IntOption windowSizeOption = new IntOption("windowSize", 
            'w', "Window size for the reactive learner", 
            12, 1, Integer.MAX_VALUE);

    public FloatOption thresholdOption = new FloatOption("threshold",
            't', "Threashold for creating a new stable learner",
            0.2, 0, 1);

    protected int[] c;
    protected Classifier stableLearner;
    protected Classifier reactiveLearner;
    protected int t;
    protected int w;
    protected int theta;
    protected Instance[] instances;
    protected int changeDetected = 0;
    protected int numberOfErrors = 0;
    protected int i;

    @Override
    public void resetLearningImpl() {
        this.t = 0;
        this.w = this.windowSizeOption.getValue();
        this.c = new int[this.w];
        this.theta = (int) (this.w * this.thresholdOption.getValue());
        this.instances = new Instance[this.w];

        this.stableLearner = ((Classifier) getPreparedClassOption(this.stableLearnerOption)).copy();
        this.stableLearner.resetLearning();
        this.reactiveLearner = ((Classifier) getPreparedClassOption(this.reactiveLearnerOption)).copy();
        this.reactiveLearner.resetLearning();
    }

    @Override
    public void trainOnInstanceImpl(Instance inst) {
        this.instances[this.t] = inst;
        int trueClass = (int) inst.classValue();
        boolean stablePrediction = MiscUtils.maxIndex(this.stableLearner.getVotesForInstance(inst)) == trueClass;
        boolean reactivePrediction = MiscUtils.maxIndex(this.reactiveLearner.getVotesForInstance(inst)) == trueClass;

        this.numberOfErrors = this.numberOfErrors - this.c[this.t];
        if(!stablePrediction && reactivePrediction) {
            this.c[this.t] = 1;
            this.numberOfErrors++;
	} else {
            this.c[this.t] = 0;
        }        
        if (this.theta < this.numberOfErrors) {
            this.changeDetected++;
            this.stableLearner = this.reactiveLearner.copy();
            Arrays.fill(this.c, 0);   // Resets c
            this.numberOfErrors = 0;
        }
        this.stableLearner.trainOnInstance(inst);
        this.reactiveLearner.resetLearning();
        for (i=0; i measurementList = new LinkedList();
        measurementList.add(new Measurement("Change detected", this.changeDetected));
        Measurement[] modelMeasurements = ((AbstractClassifier) this.stableLearner).getModelMeasurements();
        if (modelMeasurements != null) {
            measurementList.addAll(Arrays.asList(modelMeasurements));
        }
        this.changeDetected = 0;
        return measurementList.toArray(new Measurement[measurementList.size()]);
    }
    @Override
    public void getModelDescription(StringBuilder out, int indent) {

    }
}