moa.clusterers.WekaClusteringAlgorithm 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!
/*
 *    WekaClusteringAlgorithm.java
 *    Copyright (C) 2010 RWTH Aachen University, Germany
 *    @author Jansen ([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.clusterers;

import java.util.ArrayList;
import java.util.List;
import moa.cluster.Clustering;
import moa.core.AutoClassDiscovery;
import moa.core.AutoExpandVector;
import moa.core.Measurement;
import moa.options.ClassOption;
import com.github.javacliparser.IntOption;
import com.github.javacliparser.MultiChoiceOption;
import com.github.javacliparser.StringOption;
import moa.core.FastVector;
import com.yahoo.labs.samoa.instances.Attribute;
import com.yahoo.labs.samoa.instances.DenseInstance;
import com.yahoo.labs.samoa.instances.Instance;
import com.yahoo.labs.samoa.instances.Instances;
import com.yahoo.labs.samoa.instances.SamoaToWekaInstanceConverter;

import weka.core.Utils;

public class WekaClusteringAlgorithm extends AbstractClusterer {

    private static final long serialVersionUID = 1L;
    
    public IntOption horizonOption = new IntOption("horizon",
            'h', "Range of the window.", 1000);
    
    public MultiChoiceOption wekaAlgorithmOption;
    
    public StringOption parameterOption = new StringOption("parameter", 'p',
            "Parameters that will be passed to the weka algorithm. (e.g. '-N 5' for using SimpleKmeans with 5 clusters)", "-N 5 -S 8");
    
    private Class[] clustererClasses;
    
    private Instances instances;
    
    private weka.clusterers.AbstractClusterer clusterer;
    
    protected SamoaToWekaInstanceConverter instanceConverter;
    

    public WekaClusteringAlgorithm() {
        clustererClasses = findWekaClustererClasses();
        String[] optionLabels = new String[clustererClasses.length];
        String[] optionDescriptions = new String[clustererClasses.length];

        for (int i = 0; i < clustererClasses.length; i++) {
            optionLabels[i] = clustererClasses[i].getSimpleName();
            optionDescriptions[i] = clustererClasses[i].getName();
//			We do have the parameter option info, but not really a place to show it somewhere
/*
             //System.out.println(clustererClasses[i].getSimpleName());
             for (Class c : clustererClasses[i].getInterfaces()) {
             if (c.equals(weka.core.OptionHandler.class)) {
             try {
             Enumeration options = ((weka.core.OptionHandler)clustererClasses[i].newInstance()).listOptions();
             while(options.hasMoreElements()){
             weka.core.Option o = (weka.core.Option)options.nextElement(); 
             System.out.print(o.synopsis()+" ");	
             } 
							
             } catch (InstantiationException e) {
             e.printStackTrace();
             } catch (IllegalAccessException e) {
             e.printStackTrace();
             }
			        	
             }
             }
             */
        }

        if (clustererClasses != null && clustererClasses.length > 0) {
            wekaAlgorithmOption = new MultiChoiceOption("clusterer", 'w',
                    "Weka cluster algorithm to use.",
                    optionLabels, optionDescriptions, 6);
        } else {
            horizonOption = null;
            parameterOption = null;

        }

    }

    @Override
    public void resetLearningImpl() {
        try {
            instances = null;
            String clistring = clustererClasses[wekaAlgorithmOption.getChosenIndex()].getName();
            clusterer = (weka.clusterers.AbstractClusterer) ClassOption.cliStringToObject(clistring, weka.clusterers.Clusterer.class, null);

            String rawOptions = parameterOption.getValue();
            String[] options = rawOptions.split(" ");
            if (clusterer instanceof weka.core.OptionHandler) {
                ((weka.core.OptionHandler) clusterer).setOptions(options);
                Utils.checkForRemainingOptions(options);
            }
        } catch (Exception e) {
            e.printStackTrace();
        }
        this.instanceConverter = new SamoaToWekaInstanceConverter();
    }

    @Override
    public void trainOnInstanceImpl(Instance inst) {
        if (instances == null) {
            instances = getDataset(inst.numAttributes(), 0);
        }
        instances.add(inst);
    }

    public Clustering getClusteringResult() {
        Clustering clustering = null;
        weka.core.Instances wekaInstances= this.instanceConverter.wekaInstances(instances);
        try {
            
            clusterer.buildClusterer(wekaInstances);
            int numClusters = clusterer.numberOfClusters();
            Instances dataset = getDataset(instances.numAttributes(), numClusters);
            List newInstances = new ArrayList() ; //Instances(dataset);

            for (int i = 0; i < wekaInstances.numInstances(); i++) {
                weka.core.Instance inst = wekaInstances.get(i);
                int cnum = clusterer.clusterInstance(inst);

                Instance newInst = new DenseInstance(instances.instance(cnum));
                newInst.insertAttributeAt(inst.numAttributes());
                newInst.setDataset(dataset);
                newInst.setClassValue(cnum);
                newInstances.add(newInst);
            }
            clustering = new Clustering(newInstances);

        } catch (Exception e) {
            e.printStackTrace();
        }
        instances = null;

        return clustering;
    }

    public Instances getDataset(int numdim, int numclass) {
        FastVector attributes = new FastVector();
        for (int i = 0; i < numdim; i++) {
            attributes.addElement(new Attribute("att" + (i + 1)));
        }

        if (numclass > 0) {
            FastVector classLabels = new FastVector();
            for (int i = 0; i < numclass; i++) {
                classLabels.addElement("class" + (i + 1));
            }
            attributes.addElement(new Attribute("class", classLabels));
        }

        Instances myDataset = new Instances("horizion", attributes, 0);
        if (numclass > 0) {
            myDataset.setClassIndex(myDataset.numAttributes() - 1);
        }

        return myDataset;
    }

    private Class[] findWekaClustererClasses() {
        AutoExpandVector> finalClasses = new AutoExpandVector>();
        Class[] classesFound = AutoClassDiscovery.findClassesOfType("weka.clusterers",
                weka.clusterers.AbstractClusterer.class);
        for (Class foundClass : classesFound) {
            finalClasses.add(foundClass);
        }
        return finalClasses.toArray(new Class[finalClasses.size()]);
    }

    @Override
    protected Measurement[] getModelMeasurementsImpl() {
        throw new UnsupportedOperationException("Not supported yet.");
    }

    @Override
    public void getModelDescription(StringBuilder out, int indent) {
    }

    public boolean isRandomizable() {
        return false;
    }

    public double[] getVotesForInstance(Instance inst) {
        return null;
    }

    @Override
    public boolean keepClassLabel() {
        return false;
    }

    @Override
    public String getPurposeString() {
        String purpose = "MOA Clusterer: " + getClass().getCanonicalName();
        if (clustererClasses == null || clustererClasses.length == 0) {
            purpose += "\nPlease add weka.jar to the classpath to use Weka clustering algorithms.";
        }
        return purpose;

    }
}