corpusbasedfeatureweighting.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.features.weighting;

import java.util.*;

import edu.udo.cs.yale.example.*;
import edu.udo.cs.yale.operator.*;
import edu.udo.cs.yale.operator.parameter.*;

/** This operator uses a corpus of examples to characterize a single class by setting feature weights.
 *  Characteristic features receive higher weights than less characteristic features. The weight for 
 *  a feature is determined by calculating the average value of this feature for all examples of the 
 *  target class. This operator assumes that the feature values characterize the importance of this feature 
 *  for an example (e.g. TFIDF or others). Therefore, this operator is mainly used on textual data based on
 *  TFIDF weighting schemes. To extract such feature values from text collections you can use the Word Vector
 *  Tool plugin.
 *
 *  @author Michael Wurst, Ingo Mierswa
 *  @version $Id: CorpusBasedFeatureWeighting.java,v 1.2 2004/09/17 16:47:16 ingomierswa Exp $
 */
public class CorpusBasedFeatureWeighting extends Operator{

	private static String TARGET_VALUE_NAME = "class_to_characterize";

	private double epsilon = 0.001;

	/**
	 * @see edu.udo.cs.yale.operator.Operator#apply()
	 */
	public IOObject[] apply() throws OperatorException {
		
		ExampleSet es = (ExampleSet) getInput(ExampleSet.class, false);
		
		String targetValue = getParameterAsString(TARGET_VALUE_NAME);
		
		double[] weights = generateWeightsForClass(es, targetValue);
		double maxWeight = Double.NEGATIVE_INFINITY;
		for (int i = 0; i < weights.length; i++)
		    if (weights[i] > maxWeight)
			maxWeight = weights[i];
		maxWeight += epsilon;

		AttributeWeights attWeights = new AttributeWeights();
		
		for(int i = 0; i < es.getNumberOfAttributes(); i++)
			if(weights[i] > 0.0)
				attWeights.setWeight(es.getAttribute(i).getName(), weights[i]/ maxWeight);
			else
				attWeights.setWeight(es.getAttribute(i).getName(), -1.0);	
		
		return new IOObject[]{attWeights};
	}
 
	/**
	 * @see edu.udo.cs.yale.operator.Operator#getInputClasses()
	 */
	public Class[] getInputClasses() {
		return new Class[]{ExampleSet.class};
	}

	/**
	 * @see edu.udo.cs.yale.operator.Operator#getOutputClasses()
	 */
	public Class[] getOutputClasses() {
		return new Class[]{ExampleSet.class, AttributeWeights.class};
	}

	private double[] generateWeightsForClass(ExampleSet es, String value) {
		
		double[] result = new double[es.getNumberOfAttributes()];
		for(int i = 0; i < es.getNumberOfAttributes(); i++) 
			result[i] = 0.0;
		
		ExampleReader er = es.getExampleReader();
		
		while(er.hasNext()) {
			
			Example e = er.next();	
			if(e.getLabelAsString().equalsIgnoreCase(value)) {
				
				for(int i = 0; i < es.getNumberOfAttributes(); i++) 
					result[i] = result[i] + e.getValue(i);
					
			}	
			
		}
		
		return result;
		
	}


    public List getParameterTypes() {
	List types = super.getParameterTypes();
	types.add(new ParameterTypeString(TARGET_VALUE_NAME,
					  "The target class for which to find characteristic feature weights.",
					  false));
	return types;
    }
}