lmt.java

JAVA的一个程序

/*
 *    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.
 */

/*
 *    LMT.java
 *    Copyright (C) 2003 Niels Landwehr
 *
 */

package weka.classifiers.trees;

import weka.classifiers.Classifier;
import weka.classifiers.Evaluation;
import weka.classifiers.trees.j48.C45ModelSelection;
import weka.classifiers.trees.j48.ModelSelection;
import weka.classifiers.trees.lmt.LMTNode;
import weka.classifiers.trees.lmt.ResidualModelSelection;
import weka.core.AdditionalMeasureProducer;
import weka.core.Capabilities;
import weka.core.Drawable;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.TechnicalInformation;
import weka.core.TechnicalInformation.Type;
import weka.core.TechnicalInformation.Field;
import weka.core.TechnicalInformationHandler;
import weka.core.Utils;
import weka.core.Capabilities.Capability;
import weka.filters.Filter;
import weka.filters.supervised.attribute.NominalToBinary;
import weka.filters.unsupervised.attribute.ReplaceMissingValues;

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

/**
 <!-- globalinfo-start -->
 * Classifier for building 'logistic model trees', which are classification trees with logistic regression functions at the leaves. The algorithm can deal with binary and multi-class target variables, numeric and nominal attributes and missing values.<br/>
 * <br/>
 * For more information see: <br/>
 * <br/>
 * Niels Landwehr, Mark Hall, Eibe Frank (2005). Logistic Model Trees.<br/>
 * <br/>
 * Marc Sumner, Eibe Frank, Mark Hall: Speeding up Logistic Model Tree Induction. In: 9th European Conference on Principles and Practice of Knowledge Discovery in Databases, 675-683, 2005.
 * <p/>
 <!-- globalinfo-end -->
 *
 <!-- technical-bibtex-start -->
 * BibTeX:
 * <pre>
 * &#64;article{Landwehr2005,
 *    author = {Niels Landwehr and Mark Hall and Eibe Frank},
 *    booktitle = {Machine Learning},
 *    number = {1-2},
 *    pages = {161-205},
 *    title = {Logistic Model Trees},
 *    volume = {95},
 *    year = {2005}
 * }
 * 
 * &#64;inproceedings{Sumner2005,
 *    author = {Marc Sumner and Eibe Frank and Mark Hall},
 *    booktitle = {9th European Conference on Principles and Practice of Knowledge Discovery in Databases},
 *    pages = {675-683},
 *    publisher = {Springer},
 *    title = {Speeding up Logistic Model Tree Induction},
 *    year = {2005}
 * }
 * </pre>
 * <p/>
 <!-- technical-bibtex-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -B
 *  Binary splits (convert nominal attributes to binary ones)</pre>
 * 
 * <pre> -R
 *  Split on residuals instead of class values</pre>
 * 
 * <pre> -C
 *  Use cross-validation for boosting at all nodes (i.e., disable heuristic)</pre>
 * 
 * <pre> -P
 *  Use error on probabilities instead of misclassification error for stopping criterion of LogitBoost.</pre>
 * 
 * <pre> -I &lt;numIterations&gt;
 *  Set fixed number of iterations for LogitBoost (instead of using cross-validation)</pre>
 * 
 * <pre> -M &lt;numInstances&gt;
 *  Set minimum number of instances at which a node can be split (default 15)</pre>
 * 
 * <pre> -W &lt;beta&gt;
 *  Set beta for weight trimming for LogitBoost. Set to 0 (default) for no weight trimming.</pre>
 * 
 * <pre> -A
 *  The AIC is used to choose the best iteration.</pre>
 * 
 <!-- options-end -->
 *
 * @author Niels Landwehr 
 * @author Marc Sumner 
 * @version $Revision: 1.8 $
 */
public class LMT 
  extends Classifier 
  implements OptionHandler, AdditionalMeasureProducer, Drawable,
             TechnicalInformationHandler {
    
  /** for serialization */
  static final long serialVersionUID = -1113212459618104943L;
  
  /** Filter to replace missing values*/
  protected ReplaceMissingValues m_replaceMissing;
    
  /** Filter to replace nominal attributes*/
  protected NominalToBinary m_nominalToBinary;
    
  /** root of the logistic model tree*/
  protected LMTNode m_tree;
    
  /** use heuristic that determines the number of LogitBoost iterations only once in the beginning?*/
  protected boolean m_fastRegression;

  /** convert nominal attributes to binary ?*/
  protected boolean m_convertNominal;

  /** split on residuals?*/
  protected boolean m_splitOnResiduals;
    
  /**use error on probabilties instead of misclassification for stopping criterion of LogitBoost?*/
  protected boolean m_errorOnProbabilities;

  /**minimum number of instances at which a node is considered for splitting*/
  protected int m_minNumInstances;

  /**if non-zero, use fixed number of iterations for LogitBoost*/
  protected int m_numBoostingIterations;
    
  /**Threshold for trimming weights. Instances with a weight lower than this (as a percentage
   * of total weights) are not included in the regression fit.
   **/
  protected double m_weightTrimBeta;
  
  /** If true, the AIC is used to choose the best LogitBoost iteration*/
  private boolean m_useAIC = false;
  
  /**
   * Creates an instance of LMT with standard options
   */
  public LMT() {
    m_fastRegression = true;
    m_numBoostingIterations = -1;
    m_minNumInstances = 15;
    m_weightTrimBeta = 0;
    m_useAIC = false;
  }    

  /**
   * 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.MISSING_CLASS_VALUES);
    
    return result;
  }

  /**
   * Builds the classifier.
   *
   * @param data the data to train with
   * @throws Exception if classifier can't be built successfully
   */
  public void buildClassifier(Instances data) throws Exception{
	
    // can classifier handle the data?
    getCapabilities().testWithFail(data);

    // remove instances with missing class
    Instances filteredData = new Instances(data);
    filteredData.deleteWithMissingClass();
    
    //replace missing values
    m_replaceMissing = new ReplaceMissingValues();
    m_replaceMissing.setInputFormat(filteredData);	
    filteredData = Filter.useFilter(filteredData, m_replaceMissing);	
	
    //possibly convert nominal attributes globally
    if (m_convertNominal) {	    
      m_nominalToBinary = new NominalToBinary();
      m_nominalToBinary.setInputFormat(filteredData);	
      filteredData = Filter.useFilter(filteredData, m_nominalToBinary);
    }

    int minNumInstances = 2;
	
    //create ModelSelection object, either for splits on the residuals or for splits on the class value 
    ModelSelection modSelection;	
    if (m_splitOnResiduals) {
      modSelection = new ResidualModelSelection(minNumInstances);
    } else {
      modSelection = new C45ModelSelection(minNumInstances, filteredData);
    }
	
    //create tree root
    m_tree = new LMTNode(modSelection, m_numBoostingIterations, m_fastRegression, 
			 m_errorOnProbabilities, m_minNumInstances, m_weightTrimBeta, m_useAIC);
    //build tree
    m_tree.buildClassifier(filteredData);

    if (modSelection instanceof C45ModelSelection) ((C45ModelSelection)modSelection).cleanup();
  }

  /** 
   * Returns class probabilities for an instance.
   *
   * @param instance the instance to compute the distribution for
   * @return the class probabilities
   * @throws Exception if distribution can't be computed successfully
   */
  public double [] distributionForInstance(Instance instance) throws Exception {
	
    //replace missing values
    m_replaceMissing.input(instance);
    instance = m_replaceMissing.output();	
	
    //possibly convert nominal attributes
    if (m_convertNominal) {
      m_nominalToBinary.input(instance);
      instance = m_nominalToBinary.output();
    }
	
    return m_tree.distributionForInstance(instance);
  }

  /**
   * Classifies an instance.
   *
   * @param instance the instance to classify
   * @return the classification
   * @throws Exception if instance can't be classified successfully
   */
  public double classifyInstance(Instance instance) throws Exception {

    double maxProb = -1;
    int maxIndex = 0;
      
    //classify by maximum probability
    double[] probs = distributionForInstance(instance);       
    for (int j = 0; j < instance.numClasses(); j++) {
      if (Utils.gr(probs[j], maxProb)) {
	maxIndex = j;
	maxProb = probs[j];
      }
    }     
    return (double)maxIndex;      
  }    
     
  /**
   * Returns a description of the classifier.
   * 
   * @return a string representation of the classifier
   */
  public String toString() {
    if (m_tree!=null) {
      return "Logistic model tree \n------------------\n" + m_tree.toString();
    } else {
      return "No tree build";
    }
  }    
    
  /**
   * Returns an enumeration describing the available options.
   *
   * @return an enumeration of all the available options.
   */
  public Enumeration listOptions() {
    Vector newVector = new Vector(8);
    
    newVector.addElement(new Option("\tBinary splits (convert nominal attributes to binary ones)",
                                    "B", 0, "-B"));
    
    newVector.addElement(new Option("\tSplit on residuals instead of class values",
                                    "R", 0, "-R"));
    
    newVector.addElement(new Option("\tUse cross-validation for boosting at all nodes (i.e., disable heuristic)",
                                    "C", 0, "-C"));
    
    newVector.addElement(new Option("\tUse error on probabilities instead of misclassification error "+
                                    "for stopping criterion of LogitBoost.",
                                    "P", 0, "-P"));
    
    newVector.addElement(new Option("\tSet fixed number of iterations for LogitBoost (instead of using "+
                                    "cross-validation)",
                                    "I",1,"-I <numIterations>"));
    
    newVector.addElement(new Option("\tSet minimum number of instances at which a node can be split (default 15)",
                                    "M",1,"-M <numInstances>"));
    
    newVector.addElement(new Option("\tSet beta for weight trimming for LogitBoost. Set to 0 (default) for no weight trimming.",
                                    "W",1,"-W <beta>"));
    
    newVector.addElement(new Option("\tThe AIC is used to choose the best iteration.",
                                    "A", 0, "-A"));
    
    return newVector.elements();
  }
    
  /**
   * Parses a given list of options. <p/>
   * 
   <!-- options-start -->
   * Valid options are: <p/>
   * 
   * <pre> -B
   *  Binary splits (convert nominal attributes to binary ones)</pre>
   * 
   * <pre> -R
   *  Split on residuals instead of class values</pre>
   * 
   * <pre> -C
   *  Use cross-validation for boosting at all nodes (i.e., disable heuristic)</pre>
   * 
   * <pre> -P
   *  Use error on probabilities instead of misclassification error for stopping criterion of LogitBoost.</pre>
   * 
   * <pre> -I &lt;numIterations&gt;
   *  Set fixed number of iterations for LogitBoost (instead of using cross-validation)</pre>
   * 
   * <pre> -M &lt;numInstances&gt;
   *  Set minimum number of instances at which a node can be split (default 15)</pre>
   * 
   * <pre> -W &lt;beta&gt;
   *  Set beta for weight trimming for LogitBoost. Set to 0 (default) for no weight trimming.</pre>
   * 
   * <pre> -A
   *  The AIC is used to choose the best iteration.</pre>
   * 
   <!-- options-end -->
   *
   * @param options the list of options as an array of strings
   * @throws Exception if an option is not supported
   */
  public void setOptions(String[] options) throws Exception {

    setConvertNominal(Utils.getFlag('B', options));
    setSplitOnResiduals(Utils.getFlag('R', options));
    setFastRegression(!Utils.getFlag('C', options));
    setErrorOnProbabilities(Utils.getFlag('P', options));

    String optionString = Utils.getOption('I', options);
    if (optionString.length() != 0) {
      setNumBoostingIterations((new Integer(optionString)).intValue());
    }
	
    optionString = Utils.getOption('M', options);
    if (optionString.length() != 0) {
      setMinNumInstances((new Integer(optionString)).intValue());
    }

    optionString = Utils.getOption('W', options);
    if (optionString.length() != 0) {
      setWeightTrimBeta((new Double(optionString)).doubleValue());
    }
    
    setUseAIC(Utils.getFlag('A', options));        
    
    Utils.checkForRemainingOptions(options);
	
  } 
    
  /**