Package weka.classifiers.rules.part

Source Code of weka.classifiers.rules.part.PruneableDecList

/*
*    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.
*/

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

package weka.classifiers.rules.part;

import weka.classifiers.trees.j48.Distribution;
import weka.classifiers.trees.j48.ModelSelection;
import weka.classifiers.trees.j48.NoSplit;
import weka.core.Instances;
import weka.core.RevisionUtils;
import weka.core.Utils;

/**
* Class for handling a partial tree structure that
* can be pruned using a pruning set.
*
* @author Eibe Frank (eibe@cs.waikato.ac.nz)
* @version $Revision: 1.10 $
*/
public class PruneableDecList
  extends ClassifierDecList {

  /** for serialization */
  private static final long serialVersionUID = -7228103346297172921L;
 
  /**
   * Constructor for pruneable partial tree structure.
   *
   * @param toSelectLocModel selection method for local splitting model
   * @param minNum minimum number of objects in leaf
   */
  public PruneableDecList(ModelSelection toSelectLocModel,
        int minNum) {
            
    super(toSelectLocModel, minNum);
  }
 
  /**
   * Method for building a pruned partial tree.
   *
   * @throws Exception if tree can't be built successfully
   */
  public void buildRule(Instances train,
      Instances test) throws Exception {
   
    buildDecList(train, test, false);

    cleanup(new Instances(train, 0));
  }

  /**
   * Builds the partial tree with hold out set
   *
   * @throws Exception if something goes wrong
   */
  public void buildDecList(Instances train, Instances test,
         boolean leaf) throws Exception {
   
    Instances [] localTrain,localTest;
    int index,ind;
    int i,j;
    double sumOfWeights;
    NoSplit noSplit;
   
    m_train = null;
    m_isLeaf = false;
    m_isEmpty = false;
    m_sons = null;
    indeX = 0;
    sumOfWeights = train.sumOfWeights();
    noSplit = new NoSplit (new Distribution((Instances)train));
    if (leaf)
      m_localModel = noSplit;
    else
      m_localModel = m_toSelectModel.selectModel(train, test);
    m_test = new Distribution(test, m_localModel);
    if (m_localModel.numSubsets() > 1) {
      localTrain = m_localModel.split(train);
      localTest = m_localModel.split(test);
      train = null;
      test = null;
      m_sons = new ClassifierDecList [m_localModel.numSubsets()];
      i = 0;
      do {
  i++;
  ind = chooseIndex();
  if (ind == -1) {
    for (j = 0; j < m_sons.length; j++)
      if (m_sons[j] == null)
        m_sons[j] = getNewDecList(localTrain[j],localTest[j],true);
    if (i < 2) {
      m_localModel = noSplit;
      m_isLeaf = true;
      m_sons = null;
      if (Utils.eq(sumOfWeights,0))
        m_isEmpty = true;
      return;
    }
    ind = 0;
    break;
  } else
    m_sons[ind] = getNewDecList(localTrain[ind],localTest[ind],false);
      } while ((i < m_sons.length) && (m_sons[ind].m_isLeaf));
     
      // Check if all successors are leaves
      for (j = 0; j < m_sons.length; j++)
  if ((m_sons[j] == null) || (!m_sons[j].m_isLeaf))
    break;
      if (j == m_sons.length) {
  pruneEnd();
  if (!m_isLeaf)
    indeX = chooseLastIndex();
      }else
  indeX = chooseLastIndex();
    }else{
      m_isLeaf = true;
      if (Utils.eq(sumOfWeights, 0))
  m_isEmpty = true;
    }
  }
 
  /**
   * Returns a newly created tree.
   *
   * @param train train data
   * @param test test data
   * @param leaf
   * @throws Exception if something goes wrong
   */
  protected ClassifierDecList getNewDecList(Instances train, Instances test,
              boolean leaf) throws Exception {
  
    PruneableDecList newDecList =
      new PruneableDecList(m_toSelectModel, m_minNumObj);
   
    newDecList.buildDecList((Instances)train, test, leaf);
   
    return newDecList;
  }

  /**
   * Prunes the end of the rule.
   */
  protected void pruneEnd() throws Exception {
   
    double errorsLeaf, errorsTree;
   
    errorsTree = errorsForTree();
    errorsLeaf = errorsForLeaf();
    if (Utils.smOrEq(errorsLeaf,errorsTree)){
      m_isLeaf = true;
      m_sons = null;
      m_localModel = new NoSplit(localModel().distribution());
    }
  }

  /**
   * Computes error estimate for tree.
   */
  private double errorsForTree() throws Exception {

    Distribution test;

    if (m_isLeaf)
      return errorsForLeaf();
    else {
      double error = 0;
      for (int i = 0; i < m_sons.length; i++)
  if (Utils.eq(son(i).localModel().distribution().total(),0)) {
    error += m_test.perBag(i)-
      m_test.perClassPerBag(i,localModel().distribution().
        maxClass());
  } else
    error += ((PruneableDecList)son(i)).errorsForTree();

      return error;
    }
  }

  /**
   * Computes estimated errors for leaf.
   */
  private double errorsForLeaf() throws Exception {

    return m_test.total()-
      m_test.perClass(localModel().distribution().maxClass());
  }
 
  /**
   * Returns the revision string.
   *
   * @return    the revision
   */
  public String getRevision() {
    return RevisionUtils.extract("$Revision: 1.10 $");
  }
}
TOP

Related Classes of weka.classifiers.rules.part.PruneableDecList

TOP
Copyright © 2018 www.massapi.com. All rights reserved.
All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.