mioptimalball.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

   * calculate the distances from each instance in a positive bag to each bag.
   * All result distances are stored in m_Distance[i][j][k], where
   * m_Distance[i][j][k] refers the distances from the jth instance in ith bag
   * to the kth bag 
   * 
   * @param train the multi-instance dataset (with relational attribute)   
   */
  public void calculateDistance (Instances train) {
    int numBags =train.numInstances();
    int numInstances;
    Instance tempCenter;

    m_Distance = new double [numBags][][];
    for (int i=0; i<numBags; i++) {
      if (train.instance(i).classValue() == 1.0) { //positive bag
        numInstances = train.instance(i).relationalValue(1).numInstances();
        m_Distance[i]= new double[numInstances][];
        for (int j=0; j<numInstances; j++) {
          tempCenter = train.instance(i).relationalValue(1).instance(j);
          m_Distance[i][j]=new double [numBags];  //store the distance from one center to all the bags
          for (int k=0; k<numBags; k++){
            if (i==k)
              m_Distance[i][j][k]= 0;     
            else 
              m_Distance[i][j][k]= minBagDistance (tempCenter, train.instance(k));    	    
          }
        }
      } 
    }
  } 

  /**
   * Calculate the distance from one data point to a bag
   *
   * @param center the data point in instance space
   * @param bag the bag 
   * @return the double value as the distance.
   */
  public double minBagDistance (Instance center, Instance bag){
    double distance;
    double minDistance = Double.MAX_VALUE;
    Instances temp = bag.relationalValue(1);  
    //calculate the distance from the data point to each instance in the bag and return the minimum distance 
    for (int i=0; i<temp.numInstances(); i++){
      distance =0;
      for (int j=0; j<center.numAttributes(); j++)
        distance += (center.value(j)-temp.instance(i).value(j))*(center.value(j)-temp.instance(i).value(j));

      if (minDistance>distance)
        minDistance = distance;
    }
    return Math.sqrt(minDistance); 
  }

  /**
   * Find the maximum radius for the optimal ball.
   *
   * @param train the multi-instance data 
   */ 
  public void findRadius(Instances train) {
    int numBags, numInstances;
    double radius, bagDistance;
    int highestCount=0;

    numBags = train.numInstances();
    //try each instance in all positive bag as a ball center (tempCenter),   	
    for (int i=0; i<numBags; i++) {
      if (train.instance(i).classValue()== 1.0) {//positive bag   
        numInstances = train.instance(i).relationalValue(1).numInstances();
        for (int j=0; j<numInstances; j++) {   	    		
          Instance tempCenter = train.instance(i).relationalValue(1).instance(j);

          //set the possible set of ball radius corresponding to each tempCenter,
          double sortedDistance[] = sortArray(m_Distance[i][j]); //sort the distance value    	          
          for (int k=1; k<sortedDistance.length; k++){
            radius = sortedDistance[k]-(sortedDistance[k]-sortedDistance[k-1])/2.0 ;

            //evaluate the performance on the training data according to
            //the curren selected tempCenter and the set of radius   
            int correctCount =0;
            for (int n=0; n<numBags; n++){
              bagDistance=m_Distance[i][j][n];  
              if ((bagDistance <= radius && train.instance(n).classValue()==1.0) 
                  ||(bagDistance > radius && train.instance(n).classValue ()==0.0))
                correctCount += train.instance(n).weight();

            }

            //and keep the track of the ball center and the maximum radius which can achieve the highest accuracy. 
            if (correctCount > highestCount || (correctCount==highestCount && radius > m_Radius)){
              highestCount = correctCount;
              m_Radius = radius;
              for (int p=0; p<tempCenter.numAttributes(); p++)
                m_Center[p]= tempCenter.value(p);
            }      
          }
        }
      }
    } 
  }

  /**
   * Sort the array.
   *
   * @param distance the array need to be sorted
   * @return sorted array
   */ 
  public double [] sortArray(double [] distance) {
    double [] sorted = new double [distance.length];

    //make a copy of the array
    double []disCopy = new double[distance.length];
    for (int i=0;i<distance.length; i++)
      disCopy[i]= distance[i];

    DoubleVector sortVector = new DoubleVector(disCopy);
    sortVector.sort();
    sorted = sortVector.getArrayCopy(); 
    return sorted;
  }


  /**
   * Computes the distribution for a given multiple instance
   *
   * @param newBag the instance for which distribution is computed
   * @return the distribution
   * @throws Exception if the distribution can't be computed successfully
   */
  public double[] distributionForInstance(Instance newBag)
    throws Exception {  

    double [] distribution = new double[2];	
    double distance; 
    distribution[0]=0;	 
    distribution[1]=0;

    Instances insts = new Instances(newBag.dataset(),0);
    insts.add(newBag);  

    // Filter instances 
    insts= Filter.useFilter( insts, m_ConvertToSI); 	
    if (m_Filter!=null) 
      insts = Filter.useFilter(insts, m_Filter);     

    //calculate the distance from each single instance to the ball center
    int numInsts = insts.numInstances(); 		
    insts.deleteAttributeAt(0); //remove the bagIndex attribute, no use for the distance calculation

    for (int i=0; i<numInsts; i++){
      distance =0;	   
      for (int j=0; j<insts.numAttributes()-1; j++)
        distance += (insts.instance(i).value(j) - m_Center[j])*(insts.instance(i).value(j)-m_Center[j]);  

      if (distance <=m_Radius*m_Radius){  // check whether this single instance is inside the ball
        distribution[1]=1.0;  //predicted as a positive bag   	  
        break;
      }	
    }

    distribution[0]= 1-distribution[1]; 

    return distribution; 
  }

  /**
   * Returns an enumeration describing the available options.
   *
   * @return an enumeration of all the available options.
   */
  public Enumeration listOptions() {
    Vector result = new Vector();

    result.addElement(new Option(
          "\tWhether to 0=normalize/1=standardize/2=neither. \n"
          + "\t(default 0=normalize)",
          "N", 1, "-N <num>"));

    return result.elements();
  }

  /**
   * Gets the current settings of the classifier.
   *
   * @return an array of strings suitable for passing to setOptions
   */
  public String[] getOptions() {
    Vector        result;
    
    result = new Vector();

    if (getDebug())
      result.add("-D");
    
    result.add("-N");
    result.add("" + m_filterType);

    return (String[]) result.toArray(new String[result.size()]);
  }

  /**
   * Parses a given list of options. <p/>
   * 
   <!-- options-start -->
   * Valid options are: <p/>
   * 
   * <pre> -N &lt;num&gt;
   *  Whether to 0=normalize/1=standardize/2=neither. 
   *  (default 0=normalize)</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 {
    setDebug(Utils.getFlag('D', options));

    String nString = Utils.getOption('N', options);
    if (nString.length() != 0) {
      setFilterType(new SelectedTag(Integer.parseInt(nString), TAGS_FILTER));
    } else {
      setFilterType(new SelectedTag(FILTER_NORMALIZE, TAGS_FILTER));
    }
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String filterTypeTipText() {
    return "The filter type for transforming the training data.";
  }

  /**
   * Sets how the training data will be transformed. Should be one of
   * FILTER_NORMALIZE, FILTER_STANDARDIZE, FILTER_NONE.
   *
   * @param newType the new filtering mode
   */
  public void setFilterType(SelectedTag newType) {

    if (newType.getTags() == TAGS_FILTER) {
      m_filterType = newType.getSelectedTag().getID();
    }
  }

  /**
   * Gets how the training data will be transformed. Will be one of
   * FILTER_NORMALIZE, FILTER_STANDARDIZE, FILTER_NONE.
   *
   * @return the filtering mode
   */
  public SelectedTag getFilterType() {

    return new SelectedTag(m_filterType, TAGS_FILTER);
  }

  /**
   * Main method for testing this class.
   *
   * @param argv should contain the command line arguments to the
   * scheme (see Evaluation)
   */
  public static void main(String[] argv) {
    runClassifier(new MIOptimalBall(), argv);
  }
}