wekaclassifier.java

一个很好的LIBSVM的JAVA源码。对于要研究和改进SVM算法的学者。可以参考。来自数据挖掘工具YALE工具包。

/*
 *  YALE - Yet Another Learning Environment
 *  Copyright (C) 2001-2004
 *      Simon Fischer, Ralf Klinkenberg, Ingo Mierswa, 
 *          Katharina Morik, Oliver Ritthoff
 *      Artificial Intelligence Unit
 *      Computer Science Department
 *      University of Dortmund
 *      44221 Dortmund,  Germany
 *  email: [e-mail]
 *  web:   http://yale.cs.uni-dortmund.de/
 *
 *  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 2 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, write to the Free Software
 *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307
 *  USA.
 */
package edu.udo.cs.yale.operator.learner.weka;

import edu.udo.cs.yale.operator.learner.SerializableModel;
import edu.udo.cs.yale.operator.OperatorException;
import edu.udo.cs.yale.example.Attribute;
import edu.udo.cs.yale.example.Example;
import edu.udo.cs.yale.example.ExampleSet;
import edu.udo.cs.yale.example.ExampleReader;
import edu.udo.cs.yale.tools.LogService;
import edu.udo.cs.yale.tools.Ontology;
import edu.udo.cs.yale.tools.WekaTools;

import weka.core.Instances;
import weka.core.Instance;
import weka.core.Drawable;
import weka.gui.treevisualizer.TreeDisplayListener;
import weka.gui.treevisualizer.TreeDisplayEvent;
import weka.gui.treevisualizer.PlaceNode2;
import weka.gui.treevisualizer.TreeVisualizer;
import weka.classifiers.Classifier;

import java.awt.Component;

/** A Weka {@link weka.classifiers.Classifier} which can be used to classify
 *  {@link Example}s. It is learned by a {@link WekaLearner}.
 *
 *  @author ingo 
 *  @version $Id: WekaClassifier.java,v 1.8 2004/08/27 11:57:42 ingomierswa Exp $
 */
public class WekaClassifier extends SerializableModel {

    /** The used weka classifier. */
    private Classifier classifier;

    /** Set to true iff this classifier is a distribution classifier which should deliver a 
     *  distribution instead of a classification value. The predicted label should be the confidence
     *  (and not the class index). */
    private boolean useDistributionClassifier = false;

    protected WekaClassifier() { super(); }

    public WekaClassifier(Attribute label) {
	super(label);
    }

    public WekaClassifier(Attribute label, Classifier classifier) {
	this(label, classifier, false);
    }

    public WekaClassifier(Attribute label, Classifier classifier, boolean setConfidence) {
	super(label);
	this.classifier = classifier;
	this.useDistributionClassifier = setConfidence;
    }

    /** Returns true iff the parameter use_distribution was set and this classifier is a distribution classifier. */
    public boolean isDistributionClassifier() {
	return useDistributionClassifier;
    }

    public void apply(ExampleSet exampleSet) throws OperatorException {
	LogService.logMessage("Converting to Weka instances.", LogService.MINIMUM);
	Attribute predictedLabel = exampleSet.getPredictedLabel();
	Instances instances = WekaTools.toWekaInstances(exampleSet, 
							"ApplierInstances", 
							predictedLabel, 
							false);
	LogService.logMessage("Applying Weka classifier.", LogService.MINIMUM);
	int i = 0;
	ExampleReader r = exampleSet.getExampleReader();
	while (r.hasNext()) {
	    Example e = r.next();
	    Instance instance = instances.instance(i++);
	    applyModelForInstance(instance, e, predictedLabel);
	}
    }

    /** Classifies ervery weka instance and sets the result as predicted label of the current example.
     */
    public void applyModelForInstance(Instance instance, Example e, Attribute predictedLabelAttribute) {
	double predictedLabel = Double.NaN;
	try {
	    if (useDistributionClassifier) {
		double confidence[] = classifier.distributionForInstance(instance); 
		// TODO: to check
		predictedLabel = 1 - confidence[0];
	    } else {
		double wekaPrediction = classifier.classifyInstance(instance);
		if (predictedLabelAttribute.isNominal()) {
		    String classification = instance.classAttribute().value((int)wekaPrediction);
		    predictedLabel = predictedLabelAttribute.mapString(classification);
		} else {
		    predictedLabel = wekaPrediction;
		}
	    }
	} catch (Exception exc) {
	    LogService.logMessage("Exception occured while classifying example:"+exc.getMessage(), 
				  LogService.ERROR);
	}
	e.setPredictedLabel(predictedLabel);
    }   

    public String toString() {
	return 
	    "Weka model ("+classifier.getClass().getName()+") for label " + getLabel() + "\n" + 
	    classifier.toString();
    }

    public String toResultString() {
	return classifier.toString();
    }

    public Component getVisualisationComponent() {
	if (classifier instanceof Drawable) {
	    try {
		Drawable drawable = (Drawable)classifier;
		return new TreeVisualizer(new TreeDisplayListener() {
			public void userCommand(TreeDisplayEvent e) {
			    //System.out.println("TreeDisplayEvent: "+e);
			}
		    },
					  drawable.graph(),
					  new PlaceNode2());
	    } catch (Exception e) {
		e.printStackTrace();
		return super.getVisualisationComponent();
	    }
	} else {
	    return super.getVisualisationComponent();
	}
    }

    public Attribute createPredictedLabel(ExampleSet exampleSet, String name) {
	Attribute predictedLabel = super.createPredictedLabel(exampleSet, name);
	if (isDistributionClassifier()) {
	    if (predictedLabel.isNominal())
		predictedLabel.clearMaps();
	    predictedLabel.setValueType(Ontology.REAL);
	}
	return predictedLabel;
    }

    public boolean equals(Object o) {
	if (!super.equals(o)) return false;
	WekaClassifier other = (WekaClassifier)o;
	if (other.useDistributionClassifier != this.useDistributionClassifier) return false;
	if (!other.classifier.getClass().equals(this.classifier.getClass())) return false;
	return true;
    }

}