graftsplit.java

来自「Weka」· Java 代码 · 共 541 行 · 第 1/2 页

   */
  public int subsetOfInterest() {
    if(m_testType == 2)
       return 0;
    if(m_testType == 3)
       return 1;
    return m_testType;
  }


  /**
   * @return the number of positive cases in the subset of interest
   */
  public double positivesForSubsetOfInterest() {
    return (m_distribution.perClassPerBag(subsetOfInterest(), m_maxClass));
  }


  /**
   * @param subset the subset to get the positives for
   * @return the number of positive cases in the specified subset
   */
  public double positives(int subset) {
    return (m_distribution.perClassPerBag(subset, 
                                    m_distribution.maxClass(subset)));
  }


  /**
   * @return the number of instances in the subset of interest
   */
  public double totalForSubsetOfInterest() {
    return (m_distribution.perBag(subsetOfInterest()));
  }

  
  /**
   * @param subset the index of the bag to get the total for
   * @return the number of instances in the subset
   */
  public double totalForSubset(int subset) {
    return (m_distribution.perBag(subset));
  }


  /**
   * Prints left side of condition satisfied by instances.
   *
   * @param data the data.
   */
  public String leftSide(Instances data) {
    return data.attribute(m_attIndex).name();
  }


  /**
   * @return the index of the attribute to split on
   */ 
  public int attribute() {
    return m_attIndex;
  }


  /**
   * Prints condition satisfied by instances in subset index.
   */
  public final String rightSide(int index, Instances data) {

    StringBuffer text;

    text = new StringBuffer();
    if(data.attribute(m_attIndex).isNominal())
       if(index == 0)
          text.append(" = "+
                      data.attribute(m_attIndex).value((int)m_splitPoint));
       else
          text.append(" != "+
                      data.attribute(m_attIndex).value((int)m_splitPoint));
    else
       if(index == 0)
          text.append(" <= "+
                      Utils.doubleToString(m_splitPoint,6));
       else
          text.append(" > "+
                      Utils.doubleToString(m_splitPoint,6));
    return text.toString();
  }


  /**
   * Returns a string containing java source code equivalent to the test
   * made at this node. The instance being tested is called "i".
   *
   * @param index index of the nominal value tested
   * @param data the data containing instance structure info
   * @return a value of type 'String'
   */
  public final String sourceExpression(int index, Instances data) {

    StringBuffer expr = null;
    if(index < 0) {
       return "i[" + m_attIndex + "] == null";
    }
    if(data.attribute(m_attIndex).isNominal()) {
       if(index == 0)
          expr = new StringBuffer("i[");
       else
          expr = new StringBuffer("!i[");
       expr.append(m_attIndex).append("]");
       expr.append(".equals(\"").append(data.attribute(m_attIndex)
                                      .value((int)m_splitPoint)).append("\")");
    } else {
       expr = new StringBuffer("((Double) i[");
       expr.append(m_attIndex).append("])");
       if(index == 0) {
          expr.append(".doubleValue() <= ").append(m_splitPoint);
       } else {
          expr.append(".doubleValue() > ").append(m_splitPoint);
       }
    }
    return expr.toString();
  }


  /**
   * @param instance the instance to produce the weights for
   * @return a double array of weights, null if only belongs to one subset
   */
  public double [] weights(Instance instance) {

    double [] weights;
    int i;

    if(instance.isMissing(m_attIndex)) {
       weights = new double [m_numSubsets];
       for(i=0;i<m_numSubsets;i++) {
          weights [i] = m_graftdistro.perBag(i)/m_graftdistro.total();
       }
       return weights;
    } else {
       return null;
    }
  }


  /**
   * @param instance the instance for which to determine the subset
   * @return an int indicating the subset this instance belongs to
   */
  public int whichSubset(Instance instance) {

    if(instance.isMissing(m_attIndex))
       return -1;

    if(instance.attribute(m_attIndex).isNominal()) {
       // in the case of nominal, m_splitPoint is the = value, all else is !=
       if(instance.value(m_attIndex) == m_splitPoint)
          return 0;
       else
          return 1;
    } else {
       if(Utils.smOrEq(instance.value(m_attIndex), m_splitPoint))
          return 0;
       else
          return 1;
    }
  }


  /**
   * @return the value of the split point
   */
  public double splitPoint() {
    return m_splitPoint;
  }

  /**
   * @return the dominate class for the subset of interest
   */
  public int maxClassForSubsetOfInterest() {
    return m_maxClass;
  }

  /**
   * @return the laplace value for maxClass of subset of interest
   */
  public double laplaceForSubsetOfInterest() {
    return m_laplace;
  }

  /**
   * returns the test type
   * @return value of testtype
   */
  public int testType() {
    return m_testType;
  }

  /**
   * method needed for sorting a collection of GraftSplits by laplace value
   * @param g the graft split to compare to this one
   * @return -1, 0, or 1 if this GraftSplit laplace is <, = or > than that of g
   */
  public int compareTo(Object g) {

    if(m_laplace > ((GraftSplit)g).laplaceForSubsetOfInterest())
       return 1;
    if(m_laplace < ((GraftSplit)g).laplaceForSubsetOfInterest())
       return -1;
    return 0;
  }

  /**
   * returns the probability for instance for the specified class
   * @param classIndex the index of the class
   * @param instance the instance to get the probability for
   * @param theSubset the subset
   */
  public final double classProb(int classIndex, Instance instance, 
		            int theSubset) throws Exception {

    if (theSubset <= -1) {
       double [] weights = weights(instance);
       if (weights == null) {
          return m_distribution.prob(classIndex);
       } else {
          double prob = 0;
          for (int i = 0; i < weights.length; i++) {
             prob += weights[i] * m_distribution.prob(classIndex, i);
          }
          return prob;
       }
    } else {
       if (Utils.gr(m_distribution.perBag(theSubset), 0)) {
          return m_distribution.prob(classIndex, theSubset);
       } else {
          return m_distribution.prob(classIndex);
       }
    }
  }


  /**
   * method for returning information about this GraftSplit
   * @param data instances for determining names of attributes and values
   * @return a string showing this GraftSplit's information
   */
  public String toString(Instances data) {

    String theTest;
    if(m_testType == 0)
       theTest = " <= ";
    else if(m_testType == 1)
       theTest = " > ";
    else if(m_testType == 2)
       theTest = " = ";
    else
       theTest = " != ";

    if(data.attribute(m_attIndex).isNominal())
       theTest += data.attribute(m_attIndex).value((int)m_splitPoint);
    else
       theTest += Double.toString(m_splitPoint);

    return data.attribute(m_attIndex).name() + theTest
           + " (" + Double.toString(m_laplace) + ") --> " 
           + data.attribute(data.classIndex()).value(m_maxClass);
  }
}