kstar.java

代码是一个分类器的实现,其中使用了部分weka的源代码。可以将项目导入eclipse运行

/*
 *    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.
 */

/*
 *    KStar.java
 *    Copyright (c) 1995-97 by Len Trigg ([e-mail]).
 *    Java port to Weka by Abdelaziz Mahoui ([e-mail]).
 *
 */

package weka.classifiers.lazy;

import weka.classifiers.Classifier;
import weka.classifiers.UpdateableClassifier;
import weka.classifiers.lazy.kstar.KStarCache;
import weka.classifiers.lazy.kstar.KStarConstants;
import weka.classifiers.lazy.kstar.KStarNominalAttribute;
import weka.classifiers.lazy.kstar.KStarNumericAttribute;
import weka.core.Attribute;
import weka.core.Capabilities;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Option;
import weka.core.SelectedTag;
import weka.core.Tag;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformationHandler;
import weka.core.Utils;
import weka.core.Capabilities.Capability;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformation.Type;

import java.util.Enumeration;
import java.util.Random;
import java.util.Vector;

/**
 <!-- globalinfo-start -->
 * K* is an instance-based classifier, that is the class of a test instance is based upon the class of those training instances similar to it, as determined by some similarity function.  It differs from other instance-based learners in that it uses an entropy-based distance function.<br/>
 * <br/>
 * For more information on K*, see<br/>
 * <br/>
 * John G. Cleary, Leonard E. Trigg: K*: An Instance-based Learner Using an Entropic Distance Measure. In: 12th International Conference on Machine Learning, 108-114, 1995.
 * <p/>
 <!-- globalinfo-end -->
 *
 <!-- technical-bibtex-start -->
 * BibTeX:
 * <pre>
 * &#64;inproceedings{Cleary1995,
 *    author = {John G. Cleary and Leonard E. Trigg},
 *    booktitle = {12th International Conference on Machine Learning},
 *    pages = {108-114},
 *    title = {K*: An Instance-based Learner Using an Entropic Distance Measure},
 *    year = {1995}
 * }
 * </pre>
 * <p/>
 <!-- technical-bibtex-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -B &lt;num&gt;
 *  Manual blend setting (default 20%)
 * </pre>
 * 
 * <pre> -E
 *  Enable entropic auto-blend setting (symbolic class only)
 * </pre>
 * 
 * <pre> -M &lt;char&gt;
 *  Specify the missing value treatment mode (default a)
 *  Valid options are: a(verage), d(elete), m(axdiff), n(ormal)
 * </pre>
 * 
 <!-- options-end -->
 *
 * @author Len Trigg ([e-mail])
 * @author Abdelaziz Mahoui ([e-mail])
 * @version $Revision: 1.7 $
 */
public class KStar 
  extends Classifier
  implements KStarConstants, UpdateableClassifier, TechnicalInformationHandler {

  /** for serialization */
  static final long serialVersionUID = 332458330800479083L;
  
  /** The training instances used for classification. */
  protected Instances m_Train; 

  /** The number of instances in the dataset */
  protected int m_NumInstances;

  /** The number of class values */
  protected int m_NumClasses;

  /** The number of attributes */
  protected int m_NumAttributes;

  /** The class attribute type */
  protected int m_ClassType;

  /** Table of random class value colomns */
  protected int [][] m_RandClassCols;

  /** Flag turning on and off the computation of random class colomns */
  protected int m_ComputeRandomCols = ON;

  /** Flag turning on and off the initialisation of config variables */
  protected int m_InitFlag = ON;

  /**
   * A custom data structure for caching distinct attribute values
   * and their scale factor or stop parameter.
   */
  protected KStarCache [] m_Cache;

  /** missing value treatment */
  protected int m_MissingMode = M_AVERAGE;

  /** 0 = use specified blend, 1 = entropic blend setting */
  protected int m_BlendMethod = B_SPHERE;

  /** default sphere of influence blend setting */
  protected int m_GlobalBlend = 20;

  /** Define possible missing value handling methods */
  public static final Tag [] TAGS_MISSING = {
    new Tag(M_DELETE, "Ignore the instances with missing values"),
    new Tag(M_MAXDIFF, "Treat missing values as maximally different"),
    new Tag(M_NORMAL, "Normalize over the attributes"),
    new Tag(M_AVERAGE, "Average column entropy curves")
      };
    
  /**
   * Returns a string describing classifier
   * @return a description suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {

    return "K* is an instance-based classifier, that is the class of a test "
      + "instance is based upon the class of those training instances "
      + "similar to it, as determined by some similarity function.  It differs "
      + "from other instance-based learners in that it uses an entropy-based "
      + "distance function.\n\n"
      + "For more information on K*, see\n\n"
      + getTechnicalInformation().toString();
  }

  /**
   * Returns an instance of a TechnicalInformation object, containing 
   * detailed information about the technical background of this class,
   * e.g., paper reference or book this class is based on.
   * 
   * @return the technical information about this class
   */
  public TechnicalInformation getTechnicalInformation() {
    TechnicalInformation 	result;
    
    result = new TechnicalInformation(Type.INPROCEEDINGS);
    result.setValue(Field.AUTHOR, "John G. Cleary and Leonard E. Trigg");
    result.setValue(Field.TITLE, "K*: An Instance-based Learner Using an Entropic Distance Measure");
    result.setValue(Field.BOOKTITLE, "12th International Conference on Machine Learning");
    result.setValue(Field.YEAR, "1995");
    result.setValue(Field.PAGES, "108-114");
    
    return result;
  }

  /**
   * Returns default capabilities of the classifier.
   *
   * @return      the capabilities of this classifier
   */
  public Capabilities getCapabilities() {
    Capabilities result = super.getCapabilities();

    // attributes
    result.enable(Capability.NOMINAL_ATTRIBUTES);
    result.enable(Capability.NUMERIC_ATTRIBUTES);
    result.enable(Capability.DATE_ATTRIBUTES);
    result.enable(Capability.MISSING_VALUES);

    // class
    result.enable(Capability.NOMINAL_CLASS);
    result.enable(Capability.NUMERIC_CLASS);
    result.enable(Capability.DATE_CLASS);
    result.enable(Capability.MISSING_CLASS_VALUES);

    // instances
    result.setMinimumNumberInstances(0);
    
    return result;
  }

  /**
   * Generates the classifier.
   *
   * @param instances set of instances serving as training data 
   * @throws Exception if the classifier has not been generated successfully
   */
  public void buildClassifier(Instances instances) throws Exception {
    String debug = "(KStar.buildClassifier) ";

    // can classifier handle the data?
    getCapabilities().testWithFail(instances);

    // remove instances with missing class
    instances = new Instances(instances);
    instances.deleteWithMissingClass();
    
    m_Train = new Instances(instances, 0, instances.numInstances());

    // initializes class attributes ** java-speaking! :-) **
    init_m_Attributes();
  }
  
  /**
   * Adds the supplied instance to the training set
   *
   * @param instance the instance to add
   * @throws Exception if instance could not be incorporated successfully
   */
  public void updateClassifier(Instance instance) throws Exception {
    String debug = "(KStar.updateClassifier) ";

    if (m_Train.equalHeaders(instance.dataset()) == false)
      throw new Exception("Incompatible instance types");
    if ( instance.classIsMissing() )
      return;
    m_Train.add(instance);
    // update relevant attributes ...
    update_m_Attributes();
  }

  /**
   * Calculates the class membership probabilities for the given test instance.
   *
   * @param instance the instance to be classified
   * @return predicted class probability distribution
   * @throws Exception if an error occurred during the prediction
   */
  public double [] distributionForInstance(Instance instance) throws Exception {

    String debug = "(KStar.distributionForInstance) ";
    double transProb = 0.0, temp = 0.0;
    double [] classProbability = new double[m_NumClasses];
    double [] predictedValue = new double[1];

    // initialization ...
    for (int i=0; i<classProbability.length; i++) {
      classProbability[i] = 0.0;
    }
    predictedValue[0] = 0.0;
    if (m_InitFlag == ON) {
	// need to compute them only once and will be used for all instances.
	// We are doing this because the evaluation module controls the calls. 
      if (m_BlendMethod == B_ENTROPY) {
	generateRandomClassColomns();
      }
      m_Cache = new KStarCache[m_NumAttributes];
      for (int i=0; i<m_NumAttributes;i++) {
	m_Cache[i] = new KStarCache();
      }
      m_InitFlag = OFF;
      //      System.out.println("Computing...");
    }
    // init done.
    Instance trainInstance;
    Enumeration enu = m_Train.enumerateInstances();
    while ( enu.hasMoreElements() ) {
      trainInstance = (Instance)enu.nextElement();
      transProb = instanceTransformationProbability(instance, trainInstance);      
      switch ( m_ClassType )
	{
	case Attribute.NOMINAL:
	  classProbability[(int)trainInstance.classValue()] += transProb;
	  break;
	case Attribute.NUMERIC:
	  predictedValue[0] += transProb * trainInstance.classValue();
	  temp += transProb;
	  break;
	}
    }
    if (m_ClassType == Attribute.NOMINAL) {
      double sum = Utils.sum(classProbability);
      if (sum <= 0.0)
	for (int i=0; i<classProbability.length; i++)
	  classProbability[i] = (double) 1/ (double) m_NumClasses;
      else Utils.normalize(classProbability, sum);
      return classProbability;
    }
    else {
      predictedValue[0] = (temp != 0) ? predictedValue[0] / temp : 0.0;
      return predictedValue;
    }
  }

  /**
   * Calculate the probability of the first instance transforming into the 
   * second instance:
   * the probability is the product of the transformation probabilities of 
   * the attributes normilized over the number of instances used.
   * 
   * @param first the test instance
   * @param second the train instance
   * @return transformation probability value
   */
  private double instanceTransformationProbability(Instance first, 
						   Instance second) {
    String debug = "(KStar.instanceTransformationProbability) ";
    double transProb = 1.0;
    int numMissAttr = 0;
    for (int i = 0; i < m_NumAttributes; i++) {
      if (i == m_Train.classIndex()) {
	continue; // ignore class attribute
      }
      if (first.isMissing(i)) { // test instance attribute value is missing
	numMissAttr++;
	continue;
      }
      transProb *= attrTransProb(first, second, i);
      // normilize for missing values
      if (numMissAttr != m_NumAttributes) {
	transProb = Math.pow(transProb, (double)m_NumAttributes / 
			     (m_NumAttributes - numMissAttr));
      }
      else { // weird case!
	transProb = 0.0;
      }
    }
    // normilize for the train dataset
     return transProb / m_NumInstances;
  }

  /**
   * Calculates the transformation probability of the indexed test attribute 
   * to the indexed train attribute.
   *