costsensitiveclassifier.java

MacroWeka扩展了著名数据挖掘工具weka

/*
 *    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.
 */

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

package weka.classifiers.meta;

import weka.classifiers.Classifier;
import weka.classifiers.CostMatrix;
import weka.classifiers.Evaluation;
import weka.classifiers.RandomizableSingleClassifierEnhancer;
import weka.classifiers.rules.ZeroR;
import java.io.BufferedReader;
import java.io.File;
import java.io.FileReader;
import java.io.StringReader;
import java.io.StringWriter;
import java.util.Enumeration;
import java.util.Random;
import java.util.Vector;
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.Utils;
import weka.core.WeightedInstancesHandler;
import weka.core.Drawable;
import weka.core.UnsupportedClassTypeException;
import weka.filters.Filter;

/**
 * This metaclassifier makes its base classifier cost-sensitive. Two methods
 * can be used to introduce cost-sensitivity: reweighting training instances 
 * according to the total cost assigned to each class; or predicting the class
 * with minimum expected misclassification cost (rather than the most likely 
 * class). <p>
 *
 * Valid options are:<p>
 *
 * -M <br>
 * Minimize expected misclassification cost. 
 * (default is to reweight training instances according to costs per class)<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>
 *
 * -S seed <br>
 * Random number seed used when reweighting by resampling (default 1).<p>
 *
 * -cost-matrix matrix<br>
 * The cost matrix, specified in Matlab single line format.<p>
 *
 * Options after -- are passed to the designated classifier.<p>
 *
 * @author Len Trigg ([e-mail])
 * @version $Revision: 1.1 $
 */
public class CostSensitiveClassifier extends RandomizableSingleClassifierEnhancer
  implements OptionHandler, Drawable {

  /* 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);

  /** 
   * True if the costs should be used by selecting the minimum expected
   * cost (false means weight training data by the costs)
   */
  protected boolean m_MinimizeExpectedCost;
  
  /**
   * String describing default classifier.
   */
  protected String defaultClassifierString() {
    
    return "weka.classifiers.rules.ZeroR";
  }

  /**
   * Default constructor.
   */
  public CostSensitiveClassifier() {
    m_Classifier = new weka.classifiers.rules.ZeroR();
  }

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

    Vector newVector = new Vector(5);

    newVector.addElement(new Option(
	      "\tMinimize expected misclassification cost. Default is to\n"
	      +"\treweight training instances according to costs per class",
	      "M", 0, "-M"));
    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(
              "\tThe cost matrix in Matlab single line format.",
              "cost-matrix", 1, "-cost-matrix <matrix>"));

    Enumeration enu = super.listOptions();
    while (enu.hasMoreElements()) {
      newVector.addElement(enu.nextElement());
    }

    return newVector.elements();
  }

  /**
   * Parses a given list of options. Valid options are:<p>
   *
   * -M <br>
   * Minimize expected misclassification cost.
   * (default is to reweight training instances according to costs per class)<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>
   *
   * -S seed <br>
   * Random number seed used when reweighting by resampling (default 1).<p>
   *
   * -cost-matrix matrix<br>
   * The cost matrix, specified in Matlab single line format.<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 {

    setMinimizeExpectedCost(Utils.getFlag('M', options));

    String costFile = Utils.getOption('C', options);
    if (costFile.length() != 0) {
      try {
	setCostMatrix(new CostMatrix(new BufferedReader(
				     new FileReader(costFile))));
      } catch (Exception ex) {
	// now flag as possible old format cost matrix. Delay cost matrix
	// loading until buildClassifer is called
	setCostMatrix(null);
      }
      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('D', options);
    if (demandDir.length() != 0) {
      setOnDemandDirectory(new File(demandDir));
    }

    String cost_matrix = Utils.getOption("cost-matrix", options);
    if (cost_matrix.length() != 0) {
      StringWriter writer = new StringWriter();
      CostMatrix.parseMatlab(cost_matrix).write(writer);
      setCostMatrix(new CostMatrix(new StringReader(writer.toString())));
      setCostMatrixSource(new SelectedTag(MATRIX_SUPPLIED,
                                          TAGS_MATRIX_SOURCE));
    }
    
    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 = new String [superOptions.length + 7];

    int current = 0;

    if (m_MatrixSource == MATRIX_SUPPLIED) {
      if (m_CostFile != null) {
        options[current++] = "-C";
        options[current++] = "" + m_CostFile;
      }
      else {
        options[current++] = "-cost-matrix";
        options[current++] = getCostMatrix().toMatlab();
      }
    } else {
      options[current++] = "-N";
      options[current++] = "" + getOnDemandDirectory();
    }

    if (getMinimizeExpectedCost()) {
      options[current++] = "-M";
    }

    System.arraycopy(superOptions, 0, options, current, 
		     superOptions.length);

    while (current < options.length) {
      if (options[current] == null) {
        options[current] = "";
      }
      current++;
    }

    return options;
  }

  /**
   * @return a description of the classifier suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {

    return "A metaclassifier that makes its base classifier cost-sensitive. "
      + "Two methods can be used to introduce cost-sensitivity: reweighting "
      + "training instances according to the total cost assigned to each "
      + "class; or predicting the class with minimum expected "
      + "misclassification cost (rather than the most likely class). "
      + "Performance can often be "
      + "improved by using a Bagged classifier to improve the probability "
      + "estimates of the base classifier.";
  }

  /**
   * @return tip text for this property suitable for