metacost.java

一个数据挖掘软件ALPHAMINERR的整个过程的JAVA版源代码

/*
 *    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., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 *    MetaCost.java
 *    Copyright (C) 2002 University of Waikato
 *
 */

package weka.classifiers.meta;

import java.io.BufferedReader;
import java.io.File;
import java.io.FileReader;
import java.util.Enumeration;
import java.util.Vector;

import weka.classifiers.Classifier;
import weka.classifiers.CostMatrix;
import weka.classifiers.Evaluation;
import weka.classifiers.RandomizableSingleClassifierEnhancer;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.SelectedTag;
import weka.core.Tag;
import weka.core.UnsupportedClassTypeException;
import weka.core.Utils;


/**
 * This metaclassifier makes its base classifier cost-sensitive using the
 * method specified in <p>
 *
 * Pedro Domingos (1999). <i>MetaCost: A general method for making classifiers
 * cost-sensitive</i>, Proceedings of the Fifth International Conference on 
 * Knowledge Discovery and Data Mining, pp. 155-164. Also available online at
 * <a href="http://www.cs.washington.edu/homes/pedrod/kdd99.ps.gz">
 * http://www.cs.washington.edu/homes/pedrod/kdd99.ps.gz</a>. <p>
 *
 * This classifier should produce similar results to one created by
 * passing the base learner to Bagging, which is in turn passed to a
 * CostSensitiveClassifier operating on minimum expected cost. The difference
 * is that MetaCost produces a single cost-sensitive classifier of the
 * base learner, giving the benefits of fast classification and interpretable
 * output (if the base learner itself is interpretable). This implementation 
 * uses all bagging iterations when reclassifying training data (the MetaCost
 * paper reports a marginal improvement when only those iterations containing
 * each training instance are used in reclassifying that instance). <p>
 *
 * Valid options are:<p>
 *
 * -W classname <br>
 * Specify the full class name of a classifier (required).<p>
 *
 * -C cost file <br>
 * File name of a cost matrix to use. If this is not supplied, a cost
 * matrix will be loaded on demand. The name of the on-demand file
 * is the relation name of the training data plus ".cost", and the
 * path to the on-demand file is specified with the -N option.<p>
 *
 * -N directory <br>
 * Name of a directory to search for cost files when loading costs on demand
 * (default current directory). <p>
 *
 * -I num <br>
 * Set the number of bagging iterations (default 10). <p>
 *
 * -S seed <br>
 * Random number seed used when reweighting by resampling (default 1).<p>
 *
 * -P num <br>
 * Size of each bag, as a percentage of the training size (default 100). <p>
 *
 * Options after -- are passed to the designated classifier.<p>
 *
 * @author Len Trigg ([e-mail])
 * @version $Revision$ 
 */
public class MetaCost extends RandomizableSingleClassifierEnhancer {

  /* Specify possible sources of the cost matrix */
  public static final int MATRIX_ON_DEMAND = 1;
  public static final int MATRIX_SUPPLIED = 2;
  public static final Tag [] TAGS_MATRIX_SOURCE = {
    new Tag(MATRIX_ON_DEMAND, "Load cost matrix on demand"),
    new Tag(MATRIX_SUPPLIED, "Use explicit cost matrix")
  };

  /** Indicates the current cost matrix source */
  protected int m_MatrixSource = MATRIX_ON_DEMAND;

  /** 
   * The directory used when loading cost files on demand, null indicates
   * current directory 
   */
  protected File m_OnDemandDirectory = new File(System.getProperty("user.dir"));

  /** The name of the cost file, for command line options */
  protected String m_CostFile;

  /** The cost matrix */
  protected CostMatrix m_CostMatrix = new CostMatrix(1);

  /** The number of iterations. */
  protected int m_NumIterations = 10;

  /** The size of each bag sample, as a percentage of the training size */
  protected int m_BagSizePercent = 100;
    
  /**
   * Returns a string describing classifier
   * @return a description suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {
 
    return  "This metaclassifier makes its base classifier cost-sensitive using the "
      + "method specified in\n\n"
      + "Pedro Domingos (1999) \"MetaCost: A general method for making classifiers "
      + "cost-sensitive\", Proceedings of the Fifth International Conference on "
      + "Knowledge Discovery and Data Mining, pp 155-164.\n\n"
      + "This classifier should produce similar results to one created by "
      + "passing the base learner to Bagging, which is in turn passed to a "
      + "CostSensitiveClassifier operating on minimum expected cost. The difference "
      + "is that MetaCost produces a single cost-sensitive classifier of the "
      + "base learner, giving the benefits of fast classification and interpretable "
      + "output (if the base learner itself is interpretable). This implementation  "
      + "uses all bagging iterations when reclassifying training data (the MetaCost "
      + "paper reports a marginal improvement when only those iterations containing "
      + "each training instance are used in reclassifying that instance).";
 
  }

  /**
   * Returns an enumeration describing the available options.
   *
   * @return an enumeration of all the available options.
   */
  public Enumeration listOptions() {

    Vector newVector = new Vector(6);

    newVector.addElement(new Option(
	      "\tNumber of bagging iterations.\n"
	      + "\t(default 10)",
	      "I", 1, "-I <num>"));
    newVector.addElement(new Option(
	      "\tFile name of a cost matrix to use. If this is not supplied,\n"
              +"\ta cost matrix will be loaded on demand. The name of the\n"
              +"\ton-demand file is the relation name of the training data\n"
              +"\tplus \".cost\", and the path to the on-demand file is\n"
              +"\tspecified with the -N option.",
	      "C", 1, "-C <cost file name>"));
    newVector.addElement(new Option(
              "\tName of a directory to search for cost files when loading\n"
              +"\tcosts on demand (default current directory).",
              "N", 1, "-N <directory>"));
    newVector.addElement(new Option(
              "\tSize of each bag, as a percentage of the\n" 
              + "\ttraining set size. (default 100)",
              "P", 1, "-P"));

    Enumeration em = super.listOptions();
    while (em.hasMoreElements()) {
      newVector.addElement(em.nextElement());
    }
    return newVector.elements();
  }

  /**
   * Parses a given list of options. Valid options are:<p>
   *
   * -W classname <br>
   * Specify the full class name of a classifier (required).<p>
   *
   * -C cost file <br>
   * File name of a cost matrix to use. If this is not supplied, a cost
   * matrix will be loaded on demand. The name of the on-demand file
   * is the relation name of the training data plus ".cost", and the
   * path to the on-demand file is specified with the -N option.<p>
   *
   * -N directory <br>
   * Name of a directory to search for cost files when loading costs on demand
   * (default current directory). <p>
   *
   * -I num <br>
   * Set the number of bagging iterations (default 10). <p>
   *
   * -S seed <br>
   * Random number seed used when reweighting by resampling (default 1).<p>
   *
   * -P num <br>
   * Size of each bag, as a percentage of the training size (default 100). <p>
   *
   * Options after -- are passed to the designated classifier.<p>
   *
   * @param options the list of options as an array of strings
   * @exception Exception if an option is not supported
   */
  public void setOptions(String[] options) throws Exception {

    String bagIterations = Utils.getOption('I', options);
    if (bagIterations.length() != 0) {
      setNumIterations(Integer.parseInt(bagIterations));
    } else {
      setNumIterations(10);
    }

    String bagSize = Utils.getOption('P', options);
    if (bagSize.length() != 0) {
      setBagSizePercent(Integer.parseInt(bagSize));
    } else {
      setBagSizePercent(100);
    }

    String costFile = Utils.getOption('C', options);
    if (costFile.length() != 0) {
      setCostMatrix(new CostMatrix(new BufferedReader(
                                   new FileReader(costFile))));
      setCostMatrixSource(new SelectedTag(MATRIX_SUPPLIED,
                                          TAGS_MATRIX_SOURCE));
      m_CostFile = costFile;
    } else {
      setCostMatrixSource(new SelectedTag(MATRIX_ON_DEMAND, 
                                          TAGS_MATRIX_SOURCE));
    }
    
    String demandDir = Utils.getOption('N', options);
    if (demandDir.length() != 0) {
      setOnDemandDirectory(new File(demandDir));
    }

    super.setOptions(options);
  }

  /**
   * Gets the current settings of the Classifier.
   *
   * @return an array of strings suitable for passing to setOptions
   */
  public String [] getOptions() {


    String [] superOptions = super.getOptions();
    String [] options;

    if ((m_MatrixSource == MATRIX_SUPPLIED) && (m_CostFile == null)) {
      options = new String [superOptions.length + 4];
    } else {
      options = new String [superOptions.length + 6];
    }
    int current = 0;

    if (m_MatrixSource == MATRIX_SUPPLIED) {
      if (m_CostFile != null) {
        options[current++] = "-C";
        options[current++] = "" + m_CostFile;
      }
    } else {
      options[current++] = "-N";