ibk.java

来自「Weka」· Java 代码 · 共 1,028 行 · 第 1/3 页

/*
 *    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.
 */

/*
 *    IBk.java
 *    Copyright (C) 1999 University of Waikato, Hamilton, New Zealand
 *
 */

package weka.classifiers.lazy;

import weka.classifiers.Classifier;
import weka.classifiers.UpdateableClassifier;
import weka.core.Attribute;
import weka.core.Capabilities;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.neighboursearch.LinearNNSearch;
import weka.core.neighboursearch.NearestNeighbourSearch;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.SelectedTag;
import weka.core.Tag;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformationHandler;
import weka.core.Utils;
import weka.core.WeightedInstancesHandler;
import weka.core.Capabilities.Capability;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformation.Type;
import weka.core.AdditionalMeasureProducer;

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

/**
 <!-- globalinfo-start -->
 * K-nearest neighbours classifier. Can select appropriate value of K based on cross-validation. Can also do distance weighting.<br/>
 * <br/>
 * For more information, see<br/>
 * <br/>
 * D. Aha, D. Kibler (1991). Instance-based learning algorithms. Machine Learning. 6:37-66.
 * <p/>
 <!-- globalinfo-end -->
 * 
 <!-- technical-bibtex-start -->
 * BibTeX:
 * <pre>
 * &#64;article{Aha1991,
 *    author = {D. Aha and D. Kibler},
 *    journal = {Machine Learning},
 *    pages = {37-66},
 *    title = {Instance-based learning algorithms},
 *    volume = {6},
 *    year = {1991}
 * }
 * </pre>
 * <p/>
 <!-- technical-bibtex-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -I
 *  Weight neighbours by the inverse of their distance
 *  (use when k &gt; 1)</pre>
 * 
 * <pre> -F
 *  Weight neighbours by 1 - their distance
 *  (use when k &gt; 1)</pre>
 * 
 * <pre> -K &lt;number of neighbors&gt;
 *  Number of nearest neighbours (k) used in classification.
 *  (Default = 1)</pre>
 * 
 * <pre> -E
 *  Minimise mean squared error rather than mean absolute
 *  error when using -X option with numeric prediction.</pre>
 * 
 * <pre> -W &lt;window size&gt;
 *  Maximum number of training instances maintained.
 *  Training instances are dropped FIFO. (Default = no window)</pre>
 * 
 * <pre> -X
 *  Select the number of nearest neighbours between 1
 *  and the k value specified using hold-one-out evaluation
 *  on the training data (use when k &gt; 1)</pre>
 * 
 * <pre> -A
 *  The nearest neighbour search algorithm to use (default: weka.core.neighboursearch.LinearNNSearch).
 * </pre>
 * 
 <!-- options-end -->
 *
 * @author Stuart Inglis ([e-mail])
 * @author Len Trigg ([e-mail])
 * @author Eibe Frank ([e-mail])
 * @version $Revision: 1.40 $
 */
public class IBk 
  extends Classifier 
  implements OptionHandler, UpdateableClassifier, WeightedInstancesHandler,
             TechnicalInformationHandler, AdditionalMeasureProducer {

  /** for serialization. */
  static final long serialVersionUID = -3080186098777067172L;

  /** The training instances used for classification. */
  protected Instances m_Train;

  /** The number of class values (or 1 if predicting numeric). */
  protected int m_NumClasses;

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

  /** The number of neighbours to use for classification (currently). */
  protected int m_kNN;

  /**
   * The value of kNN provided by the user. This may differ from
   * m_kNN if cross-validation is being used.
   */
  protected int m_kNNUpper;

  /**
   * Whether the value of k selected by cross validation has
   * been invalidated by a change in the training instances.
   */
  protected boolean m_kNNValid;

  /**
   * The maximum number of training instances allowed. When
   * this limit is reached, old training instances are removed,
   * so the training data is "windowed". Set to 0 for unlimited
   * numbers of instances.
   */
  protected int m_WindowSize;

  /** Whether the neighbours should be distance-weighted. */
  protected int m_DistanceWeighting;

  /** Whether to select k by cross validation. */
  protected boolean m_CrossValidate;

  /**
   * Whether to minimise mean squared error rather than mean absolute
   * error when cross-validating on numeric prediction tasks.
   */
  protected boolean m_MeanSquared;

  /** no weighting. */
  public static final int WEIGHT_NONE = 1;
  /** weight by 1/distance. */
  public static final int WEIGHT_INVERSE = 2;
  /** weight by 1-distance. */
  public static final int WEIGHT_SIMILARITY = 4;
  /** possible instance weighting methods. */
  public static final Tag [] TAGS_WEIGHTING = {
    new Tag(WEIGHT_NONE, "No distance weighting"),
    new Tag(WEIGHT_INVERSE, "Weight by 1/distance"),
    new Tag(WEIGHT_SIMILARITY, "Weight by 1-distance")
  };
  
  /** for nearest-neighbor search. */
  protected NearestNeighbourSearch m_NNSearch = new LinearNNSearch();

  /** The number of attributes the contribute to a prediction. */
  protected double m_NumAttributesUsed;
  
  /**
   * IBk classifier. Simple instance-based learner that uses the class
   * of the nearest k training instances for the class of the test
   * instances.
   *
   * @param k the number of nearest neighbors to use for prediction
   */
  public IBk(int k) {

    init();
    setKNN(k);
  }  

  /**
   * IB1 classifer. Instance-based learner. Predicts the class of the
   * single nearest training instance for each test instance.
   */
  public IBk() {

    init();
  }
  
  /**
   * Returns a string describing classifier.
   * @return a description suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {

    return  "K-nearest neighbours classifier. Can "
      + "select appropriate value of K based on cross-validation. Can also do "
      + "distance weighting.\n\n"
      + "For more information, 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.ARTICLE);
    result.setValue(Field.AUTHOR, "D. Aha and D. Kibler");
    result.setValue(Field.YEAR, "1991");
    result.setValue(Field.TITLE, "Instance-based learning algorithms");
    result.setValue(Field.JOURNAL, "Machine Learning");
    result.setValue(Field.VOLUME, "6");
    result.setValue(Field.PAGES, "37-66");
    
    return result;
  }

  /**
   * Returns the tip text for this property.
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String KNNTipText() {
    return "The number of neighbours to use.";
  }
  
  /**
   * Set the number of neighbours the learner is to use.
   *
   * @param k the number of neighbours.
   */
  public void setKNN(int k) {
    m_kNN = k;
    m_kNNUpper = k;
    m_kNNValid = false;
  }

  /**
   * Gets the number of neighbours the learner will use.
   *
   * @return the number of neighbours.
   */
  public int getKNN() {

    return m_kNN;
  }

  /**
   * Returns the tip text for this property.
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String windowSizeTipText() {
    return "Gets the maximum number of instances allowed in the training " +
      "pool. The addition of new instances above this value will result " +
      "in old instances being removed. A value of 0 signifies no limit " +
      "to the number of training instances.";
  }
  
  /**
   * Gets the maximum number of instances allowed in the training
   * pool. The addition of new instances above this value will result
   * in old instances being removed. A value of 0 signifies no limit
   * to the number of training instances.
   *
   * @return Value of WindowSize.
   */
  public int getWindowSize() {
    
    return m_WindowSize;
  }
  
  /**
   * Sets the maximum number of instances allowed in the training
   * pool. The addition of new instances above this value will result
   * in old instances being removed. A value of 0 signifies no limit
   * to the number of training instances.
   *
   * @param newWindowSize Value to assign to WindowSize.
   */
  public void setWindowSize(int newWindowSize) {
    
    m_WindowSize = newWindowSize;
  }
  
  /**
   * Returns the tip text for this property.
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String distanceWeightingTipText() {

    return "Gets the distance weighting method used.";
  }
  
  /**
   * Gets the distance weighting method used. Will be one of
   * WEIGHT_NONE, WEIGHT_INVERSE, or WEIGHT_SIMILARITY
   *
   * @return the distance weighting method used.
   */
  public SelectedTag getDistanceWeighting() {

    return new SelectedTag(m_DistanceWeighting, TAGS_WEIGHTING);
  }
  
  /**
   * Sets the distance weighting method used. Values other than
   * WEIGHT_NONE, WEIGHT_INVERSE, or WEIGHT_SIMILARITY will be ignored.
   *
   * @param newMethod the distance weighting method to use
   */
  public void setDistanceWeighting(SelectedTag newMethod) {
    
    if (newMethod.getTags() == TAGS_WEIGHTING) {
      m_DistanceWeighting = newMethod.getSelectedTag().getID();
    }
  }
  
  /**