source: branches/MetisMQI/src/main/java/weka/classifiers/bayes/NaiveBayesMultinomial.java

/*
 *    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.
 */

/*
 * NaiveBayesMultinomial.java
 * Copyright (C) 2003 University of Waikato, Hamilton, New Zealand
 */

package weka.classifiers.bayes;

import weka.classifiers.Classifier;
import weka.classifiers.AbstractClassifier;
import weka.core.Capabilities;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.RevisionUtils;
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;

/**
 <!-- globalinfo-start -->
 * Class for building and using a multinomial Naive Bayes classifier. For more information see,<br/>
 * <br/>
 * Andrew Mccallum, Kamal Nigam: A Comparison of Event Models for Naive Bayes Text Classification. In: AAAI-98 Workshop on 'Learning for Text Categorization', 1998.<br/>
 * <br/>
 * The core equation for this classifier:<br/>
 * <br/>
 * P[Ci|D] = (P[D|Ci] x P[Ci]) / P[D] (Bayes rule)<br/>
 * <br/>
 * where Ci is class i and D is a document.
 * <p/>
 <!-- globalinfo-end -->
 *
 <!-- technical-bibtex-start -->
 * BibTeX:
 * <pre>
 * &#64;inproceedings{Mccallum1998,
 *    author = {Andrew Mccallum and Kamal Nigam},
 *    booktitle = {AAAI-98 Workshop on 'Learning for Text Categorization'},
 *    title = {A Comparison of Event Models for Naive Bayes Text Classification},
 *    year = {1998}
 * }
 * </pre>
 * <p/>
 <!-- technical-bibtex-end -->
 *
 <!-- options-start -->
 * Valid options are: <p/>
 * 
 * <pre> -D
 *  If set, classifier is run in debug mode and
 *  may output additional info to the console</pre>
 * 
 <!-- options-end -->
 *
 * @author Andrew Golightly (acg4@cs.waikato.ac.nz)
 * @author Bernhard Pfahringer (bernhard@cs.waikato.ac.nz)
 * @version $Revision: 5928 $ 
 */
public class NaiveBayesMultinomial 
  extends AbstractClassifier 
  implements WeightedInstancesHandler,TechnicalInformationHandler {
  
  /** for serialization */
  static final long serialVersionUID = 5932177440181257085L;
  
  /**
   * probability that a word (w) exists in a class (H) (i.e. Pr[w|H])
   * The matrix is in the this format: probOfWordGivenClass[class][wordAttribute]
   * NOTE: the values are actually the log of Pr[w|H]
   */
  protected double[][] m_probOfWordGivenClass;
    
  /** the probability of a class (i.e. Pr[H]) */
  protected double[] m_probOfClass;
    
  /** number of unique words */
  protected int m_numAttributes;
    
  /** number of class values */
  protected int m_numClasses;
    
  /** cache lnFactorial computations */
  protected double[] m_lnFactorialCache = new double[]{0.0,0.0};
    
  /** copy of header information for use in toString method */
  protected Instances m_headerInfo;

  /**
   * Returns a string describing this classifier
   * @return a description of the classifier suitable for
   * displaying in the explorer/experimenter gui
   */
  public String globalInfo() {
    return 
        "Class for building and using a multinomial Naive Bayes classifier. "
      + "For more information see,\n\n"
      + getTechnicalInformation().toString() + "\n\n"
      + "The core equation for this classifier:\n\n"
      + "P[Ci|D] = (P[D|Ci] x P[Ci]) / P[D] (Bayes rule)\n\n"
      + "where Ci is class i and D is a document.";
  }

  /**
   * 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, "Andrew Mccallum and Kamal Nigam");
    result.setValue(Field.YEAR, "1998");
    result.setValue(Field.TITLE, "A Comparison of Event Models for Naive Bayes Text Classification");
    result.setValue(Field.BOOKTITLE, "AAAI-98 Workshop on 'Learning for Text Categorization'");
    
    return result;
  }

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

    // attributes
    result.enable(Capability.NUMERIC_ATTRIBUTES);

    // class
    result.enable(Capability.NOMINAL_CLASS);
    result.enable(Capability.MISSING_CLASS_VALUES);
    
    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 
  {
    // can classifier handle the data?
    getCapabilities().testWithFail(instances);

    // remove instances with missing class
    instances = new Instances(instances);
    instances.deleteWithMissingClass();
    
    m_headerInfo = new Instances(instances, 0);
    m_numClasses = instances.numClasses();
    m_numAttributes = instances.numAttributes();
    m_probOfWordGivenClass = new double[m_numClasses][];
        
    /*
      initialising the matrix of word counts
      NOTE: Laplace estimator introduced in case a word that does not appear for a class in the 
      training set does so for the test set
    */
    for(int c = 0; c<m_numClasses; c++)
      {
        m_probOfWordGivenClass[c] = new double[m_numAttributes];
        for(int att = 0; att<m_numAttributes; att++)
          {
            m_probOfWordGivenClass[c][att] = 1;
          }
      }
        
    //enumerate through the instances 
    Instance instance;
    int classIndex;
    double numOccurences;
    double[] docsPerClass = new double[m_numClasses];
    double[] wordsPerClass = new double[m_numClasses];
        
    java.util.Enumeration enumInsts = instances.enumerateInstances();
    while (enumInsts.hasMoreElements()) 
      {
        instance = (Instance) enumInsts.nextElement();
        classIndex = (int)instance.value(instance.classIndex());
        docsPerClass[classIndex] += instance.weight();
                
        for(int a = 0; a<instance.numValues(); a++)
          if(instance.index(a) != instance.classIndex())
            {
              if(!instance.isMissing(a))
                {
                  numOccurences = instance.valueSparse(a) * instance.weight();
                  if(numOccurences < 0)
                    throw new Exception("Numeric attribute values must all be greater or equal to zero.");
                  wordsPerClass[classIndex] += numOccurences;
                  m_probOfWordGivenClass[classIndex][instance.index(a)] += numOccurences;
                }
            } 
      }
        
    /*
      normalising probOfWordGivenClass values
      and saving each value as the log of each value
    */
    for(int c = 0; c<m_numClasses; c++)
      for(int v = 0; v<m_numAttributes; v++) 
        m_probOfWordGivenClass[c][v] = Math.log(m_probOfWordGivenClass[c][v] / (wordsPerClass[c] + m_numAttributes - 1));
        
    /*
      calculating Pr(H)
      NOTE: Laplace estimator introduced in case a class does not get mentioned in the set of 
      training instances
    */
    final double numDocs = instances.sumOfWeights() + m_numClasses;
    m_probOfClass = new double[m_numClasses];
    for(int h=0; h<m_numClasses; h++)
      m_probOfClass[h] = (double)(docsPerClass[h] + 1)/numDocs; 
  }
    
  /**
   * 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 there is a problem generating the prediction
   */
  public double [] distributionForInstance(Instance instance) throws Exception 
  {
    double[] probOfClassGivenDoc = new double[m_numClasses];
        
    //calculate the array of log(Pr[D|C])
    double[] logDocGivenClass = new double[m_numClasses];
    for(int h = 0; h<m_numClasses; h++)
      logDocGivenClass[h] = probOfDocGivenClass(instance, h);
        
    double max = logDocGivenClass[Utils.maxIndex(logDocGivenClass)];
    double probOfDoc = 0.0;
        
    for(int i = 0; i<m_numClasses; i++) 
      {
        probOfClassGivenDoc[i] = Math.exp(logDocGivenClass[i] - max) * m_probOfClass[i];
        probOfDoc += probOfClassGivenDoc[i];
      }
        
    Utils.normalize(probOfClassGivenDoc,probOfDoc);
        
    return probOfClassGivenDoc;
  }
    
  /**
   * log(N!) + (for all the words)(log(Pi^ni) - log(ni!))
   *  
   *  where 
   *      N is the total number of words
   *      Pi is the probability of obtaining word i
   *      ni is the number of times the word at index i occurs in the document
   *
   * @param inst       The instance to be classified
   * @param classIndex The index of the class we are calculating the probability with respect to
   *
   * @return The log of the probability of the document occuring given the class
   */
    
  private double probOfDocGivenClass(Instance inst, int classIndex)
  {
    double answer = 0;
    //double totalWords = 0; //no need as we are not calculating the factorial at all.
        
    double freqOfWordInDoc;  //should be double
    for(int i = 0; i<inst.numValues(); i++)
      if(inst.index(i) != inst.classIndex())
        {
          freqOfWordInDoc = inst.valueSparse(i);
          //totalWords += freqOfWordInDoc;
          answer += (freqOfWordInDoc * m_probOfWordGivenClass[classIndex][inst.index(i)] 
                     ); //- lnFactorial(freqOfWordInDoc));
        }
        
    //answer += lnFactorial(totalWords);//The factorial terms don't make 
    //any difference to the classifier's
    //accuracy, so not needed.
        
    return answer;
  }
    
  /**
   * Fast computation of ln(n!) for non-negative ints
   *
   * negative ints are passed on to the general gamma-function
   * based version in weka.core.SpecialFunctions
   *
   * if the current n value is higher than any previous one,
   * the cache is extended and filled to cover it
   *
   * the common case is reduced to a simple array lookup
   *
   * @param  n the integer 
   * @return ln(n!)
   */
    
  public double lnFactorial(int n) 
  {
    if (n < 0) return weka.core.SpecialFunctions.lnFactorial(n);
        
    if (m_lnFactorialCache.length <= n) {
      double[] tmp = new double[n+1];
      System.arraycopy(m_lnFactorialCache,0,tmp,0,m_lnFactorialCache.length);
      for(int i = m_lnFactorialCache.length; i < tmp.length; i++) 
        tmp[i] = tmp[i-1] + Math.log(i);
      m_lnFactorialCache = tmp;
    }
        
    return m_lnFactorialCache[n];
  }
    
  /**
   * Returns a string representation of the classifier.
   * 
   * @return a string representation of the classifier
   */
  public String toString()
  {
    StringBuffer result = new StringBuffer("The independent probability of a class\n--------------------------------------\n");
        
    for(int c = 0; c<m_numClasses; c++)
      result.append(m_headerInfo.classAttribute().value(c)).append("\t").append(Double.toString(m_probOfClass[c])).append("\n");
        
    result.append("\nThe probability of a word given the class\n-----------------------------------------\n\t");

    for(int c = 0; c<m_numClasses; c++)
      result.append(m_headerInfo.classAttribute().value(c)).append("\t");
        
    result.append("\n");

    for(int w = 0; w<m_numAttributes; w++)
      {
        result.append(m_headerInfo.attribute(w).name()).append("\t");
        for(int c = 0; c<m_numClasses; c++)
          result.append(Double.toString(Math.exp(m_probOfWordGivenClass[c][w]))).append("\t");
        result.append("\n");
      }

    return result.toString();
  }
  
  /**
   * Returns the revision string.
   * 
   * @return            the revision
   */
  public String getRevision() {
    return RevisionUtils.extract("$Revision: 5928 $");
  }
    
  /**
   * Main method for testing this class.
   *
   * @param argv the options
   */
  public static void main(String [] argv) {
    runClassifier(new NaiveBayesMultinomial(), argv);
  }
}