attributeinformationgain.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;

import edu.udo.cs.yale.operator.parameter.*;
import edu.udo.cs.yale.tools.LogService;
import edu.udo.cs.yale.tools.Ontology;
import edu.udo.cs.yale.example.ExampleSet;
import edu.udo.cs.yale.example.Example;
import edu.udo.cs.yale.example.ExampleReader;
import edu.udo.cs.yale.example.Tools;

import java.util.List;

/** For classification tasks the well known information gain methods derived from C4.5 (Quinlan) are used.
 *
 *  @yale.xmlclass InformationGain
 *  @author ingo
 *  @version $Id: AttributeInformationGain.java,v 2.10 2004/08/27 11:57:34 ingomierswa Exp $
 */
public class AttributeInformationGain extends Operator {

    public static final String INFORMATION_GAIN_KEY          = "attribute.information.gain";
    public static final String SMALLEST_INFORMATION_GAIN_KEY = "attribute.information.gain.smallest";

    private static final Class[] INPUT_CLASSES  = { ExampleSet.class };
    private static final Class[] OUTPUT_CLASSES = { ExampleSet.class };

    public IOObject[] apply() throws OperatorException {

  	ExampleSet eSet = (ExampleSet)getInput(ExampleSet.class);

	double[] informationGain = new double[eSet.getNumberOfAttributes()];

	if (Ontology.ATTRIBUTE_VALUE_TYPE.isA(eSet.getLabel().getValueType(), Ontology.NOMINAL)) { 

	    // RatioGain gibt an, ob bei Klassifikationsproblemen ratio gain benutzt werden soll.
	    boolean ratioGain = getParameterAsBoolean("use_ratio_gain");
	    
	    informationGain = Tools.getInformationGain(eSet, ratioGain);
	    
	} else {
	    throw new UserError(this, 101, "information gain", eSet.getLabel().getName());
	}

	normalize(informationGain);

	// Setzen der Infolabel fuer die Attribute.
	double smallestInformationGainValue = Double.POSITIVE_INFINITY;
	for (int i = 0 ; i < eSet.getNumberOfAttributes(); i++) {
	    if (informationGain[i] < smallestInformationGainValue) 
		smallestInformationGainValue = informationGain[i];
	}

	eSet.setUserData(INFORMATION_GAIN_KEY, informationGain);
	eSet.setUserData(SMALLEST_INFORMATION_GAIN_KEY, new Double(smallestInformationGainValue));
	
	return new IOObject[] { eSet };
    }
    
    /** Diese Methode uebernimmt das Normieren der Werte auf einen Bereich zwischen 0 und 1. Das informativste Attribut 
     *  bekommt dabei uebrigens den Wert 1 zugewiesen, alle anderen sind Bruchteile dieses Wertes.
     */
    public static void normalize(double[] infoGain) {
	double best = Double.NEGATIVE_INFINITY;
	for (int i = 0; i < infoGain.length; i++) 
	    if (infoGain[i] > best) best = infoGain[i];

	for (int i = 0; i < infoGain.length; i++)
	    infoGain[i] /= best;
    }


    public Class[] getOutputClasses() {
	return OUTPUT_CLASSES;
    }

    public Class[] getInputClasses() {
	return INPUT_CLASSES;
    }

    public List getParameterTypes() {
	List types = super.getParameterTypes();
	types.add(new ParameterTypeBoolean("use_ratio_gain", "If set to true the ratio gain criterion is used.", true));
	return types;
    }
}