symmetricaluncertattributeseteval.java

这是关于数据挖掘的一些算法

      if (inst.isMissing(m_classIndex)) {
        jj = nj - 1;
      }
      else {
        jj = (int)inst.value(m_classIndex);
      }

      counts[ii][jj]++;
    }

    // get the row totals
    for (i = 0; i < ni; i++) {
      sumi[i] = 0.0;

      for (j = 0; j < nj; j++) {
        //there are how many happen of a special feature value
        sumi[i] += counts[i][j];
        sum += counts[i][j];
      }
    }

    // get the column totals
    for (j = 0; j < nj; j++) {
      sumj[j] = 0.0;

      for (i = 0; i < ni; i++) {
        //a class value include how many instance.
        sumj[j] += counts[i][j];
      }
    }

    // distribute missing counts
    if (m_missing_merge &&
        (sumi[ni-1] < m_numInstances) &&
        (sumj[nj-1] < m_numInstances)) {
      double[] i_copy = new double[sumi.length];
      double[] j_copy = new double[sumj.length];
      double[][] counts_copy = new double[sumi.length][sumj.length];

      for (i = 0; i < ni; i++) {
        System.arraycopy(counts[i], 0, counts_copy[i], 0, sumj.length);
      }

      System.arraycopy(sumi, 0, i_copy, 0, sumi.length);
      System.arraycopy(sumj, 0, j_copy, 0, sumj.length);
      double total_missing = (sumi[ni - 1] + sumj[nj - 1]
                              - counts[ni - 1][nj - 1]);

      // do the missing i's
      if (sumi[ni - 1] > 0.0) { //sumi[ni - 1]: missing value contains how many values.
        for (j = 0; j < nj - 1; j++) {
          if (counts[ni - 1][j] > 0.0) {
            for (i = 0; i < ni - 1; i++) {
              temp = ((i_copy[i]/(sum - i_copy[ni - 1])) *
                      counts[ni - 1][j]);
              counts[i][j] += temp; //according to the probability of value i we distribute account of the missing degree of a class lable to it
              sumi[i] += temp;
            }

            counts[ni - 1][j] = 0.0;
          }
        }
      }

      sumi[ni - 1] = 0.0;

      // do the missing j's
      if (sumj[nj - 1] > 0.0) {
        for (i = 0; i < ni - 1; i++) {
          if (counts[i][nj - 1] > 0.0) {
            for (j = 0; j < nj - 1; j++) {
              temp = ((j_copy[j]/(sum - j_copy[nj - 1]))*counts[i][nj - 1]);
              counts[i][j] += temp;
              sumj[j] += temp;
            }

            counts[i][nj - 1] = 0.0;
          }
        }
      }

      sumj[nj - 1] = 0.0;

      // do the both missing
      if (counts[ni - 1][nj - 1] > 0.0 && total_missing != sum) {
        for (i = 0; i < ni - 1; i++) {
          for (j = 0; j < nj - 1; j++) {
            temp = (counts_copy[i][j]/(sum - total_missing)) *
              counts_copy[ni - 1][nj - 1];
            counts[i][j] += temp;
            sumi[i] += temp;
            sumj[j] += temp;
          }
        }

        counts[ni - 1][nj - 1] = 0.0;
      }
    }

    return  ContingencyTables.symmetricalUncertainty(counts);
  }

  /**
   * calculate symmetrical uncertainty between sets of attributes
   *
   * @param attributes the indexes of the attributes
   * @param classAttributes the indexes of the attributes whose combination will
   *                         be used as class label
   * @return the uncertainty
   * @throws Exception if the attribute could not be evaluated
   */
  public double evaluateAttribute (int[] attributes, int[] classAttributes)
    throws Exception {

    int i, j;     //variable for looping.
    int p;     //variable for looping.
    int ii, jj;   //specifying the position in the contingency table.
    int nnj, nni; //counting base for attributes[].
    int ni, nj;   //the nubmer of rows and columns in the ContingencyTables.

    double sum = 0.0;
    boolean b_missing_attribute = false;
    boolean b_missing_classAtrribute = false;

    if(attributes.length==0)
    {
      throw new Exception("the parameter attributes[] is empty;SEQ:W-FS-Eval-SUAS-001");
    }
    if(classAttributes.length==0)
    {
      throw new Exception("the parameter classAttributes[] is empty;SEQ:W-FS-Eval-SUAS-002");
    }

    /*calculate the number of the rows in ContingencyTable*/
    ni = m_trainInstances.attribute(attributes[0]).numValues();
    if (ni == 0)
    {
      throw new Exception("an attribute is empty;SEQ:W-FS-Eval-SUAS-003;"+1);
    }

    for (i = 1;i<attributes.length;i++)
    {
      if (m_trainInstances.attribute(attributes[i]).numValues() == 0)
      {
        throw new Exception("an attribute is empty;SEQ:W-FS-Eval-SUAS-003;" +
                            (i+1));
      }
      ni = ni*m_trainInstances.attribute(attributes[i]).numValues();
    }
    ni = ni+1;

    /*calculate the number of the colums in the ContingencyTable*/
    nj = m_trainInstances.attribute(classAttributes[0]).numValues();
    if (nj == 0)
    {
      throw new Exception("the a classAttribute is empty;SEQ:W-FS-Eval-SUAS-004;"+1);
    }

    for (i = 1;i<classAttributes.length;i++)
    {
      if (m_trainInstances.attribute(classAttributes[i]).numValues() == 0)
      {
        throw new Exception("the a classAttribute is empty;SEQ:W-FS-Eval-SUAS-004;" +
                            (i+1));
      }
      nj = nj*m_trainInstances.attribute(classAttributes[i]).numValues();
    }
    nj = nj+1;

    double[] sumi, sumj;
    Instance inst;
    double temp = 0.0;
    sumi = new double[ni];
    sumj = new double[nj];
    double[][] counts = new double[ni][nj];
    sumi = new double[ni];
    sumj = new double[nj];

    for (i = 0; i < ni; i++) {
      sumi[i] = 0.0;

      for (j = 0; j < nj; j++) {
        sumj[j] = 0.0;
        counts[i][j] = 0.0;
      }
    }

    // Fill the contingency table
    for (i = 0; i < m_numInstances; i++) {
      inst = m_trainInstances.instance(i);

      b_missing_attribute = false;
      b_missing_classAtrribute = false;

      /*get row position in contingency table*/
      nni = 1;
      ii = 0;
      for (p=attributes.length-1; p>=0; p--)
      {
        if (inst.isMissing(attributes[p])) {
           b_missing_attribute = true;
        }
        ii = ((int)inst.value(attributes[p])*nni)+ii;
        if (p<attributes.length-1){
          nni = nni * (m_trainInstances.attribute(attributes[p]).numValues());
        }
        else {
          nni = m_trainInstances.attribute(attributes[p]).numValues();
        }
      }
      if (b_missing_attribute) {
        ii = ni-1;
      }

      /*get colum position in contingency table*/
      nnj = 1;
      jj = 0;
      for (p=classAttributes.length-1; p>=0; p--)
      {
        if (inst.isMissing(classAttributes[p])) {
           b_missing_classAtrribute = true;
        }
        jj = ((int)inst.value(classAttributes[p])*nnj)+jj;
        if (p<attributes.length-1){
          nnj = nnj * (m_trainInstances.attribute(classAttributes[p]).numValues());
        }
        else {
          nnj = m_trainInstances.attribute(classAttributes[p]).numValues();
        }
      }
      if (b_missing_classAtrribute) {
        jj = nj-1;
      }

      counts[ii][jj]++;
    }

    // get the row totals
    for (i = 0; i < ni; i++) {
      sumi[i] = 0.0;

      for (j = 0; j < nj; j++) {
        //there are how many happen of a special feature value
        sumi[i] += counts[i][j];
        sum += counts[i][j];
      }
    }

    // get the column totals
    for (j = 0; j < nj; j++) {
      sumj[j] = 0.0;

      for (i = 0; i < ni; i++) {
        //a class value include how many instance.
        sumj[j] += counts[i][j];
      }
    }

    // distribute missing counts
    if (m_missing_merge &&
        (sumi[ni-1] < m_numInstances) &&
        (sumj[nj-1] < m_numInstances)) {
      double[] i_copy = new double[sumi.length];
      double[] j_copy = new double[sumj.length];
      double[][] counts_copy = new double[sumi.length][sumj.length];

      for (i = 0; i < ni; i++) {
        System.arraycopy(counts[i], 0, counts_copy[i], 0, sumj.length);
      }

      System.arraycopy(sumi, 0, i_copy, 0, sumi.length);
      System.arraycopy(sumj, 0, j_copy, 0, sumj.length);
      double total_missing = (sumi[ni - 1] + sumj[nj - 1]
                              - counts[ni - 1][nj - 1]);

      // do the missing i's
      if (sumi[ni - 1] > 0.0) { //sumi[ni - 1]: missing value contains how many values.
        for (j = 0; j < nj - 1; j++) {
          if (counts[ni - 1][j] > 0.0) {
            for (i = 0; i < ni - 1; i++) {
              temp = ((i_copy[i]/(sum - i_copy[ni - 1])) *
                      counts[ni - 1][j]);
              counts[i][j] += temp; //according to the probability of value i we distribute account of the missing degree of a class lable to it
              sumi[i] += temp;
            }

            counts[ni - 1][j] = 0.0;
          }
        }
      }

      sumi[ni - 1] = 0.0;

      // do the missing j's
      if (sumj[nj - 1] > 0.0) {
        for (i = 0; i < ni - 1; i++) {
          if (counts[i][nj - 1] > 0.0) {
            for (j = 0; j < nj - 1; j++) {
              temp = ((j_copy[j]/(sum - j_copy[nj - 1]))*counts[i][nj - 1]);
              counts[i][j] += temp;
              sumj[j] += temp;
            }

            counts[i][nj - 1] = 0.0;
          }
        }
      }

      sumj[nj - 1] = 0.0;

      // do the both missing
      if (counts[ni - 1][nj - 1] > 0.0 && total_missing != sum) {
        for (i = 0; i < ni - 1; i++) {
          for (j = 0; j < nj - 1; j++) {
            temp = (counts_copy[i][j]/(sum - total_missing)) *
              counts_copy[ni - 1][nj - 1];
            counts[i][j] += temp;
            sumi[i] += temp;
            sumj[j] += temp;
          }
        }

        counts[ni - 1][nj - 1] = 0.0;
      }
    }

    return  ContingencyTables.symmetricalUncertainty(counts);
  }


  /**
   * Return a description of the evaluator
   * @return description as a string
   */
  public String toString () {
    StringBuffer text = new StringBuffer();

    if (m_trainInstances == null) {
      text.append("\tSymmetrical Uncertainty evaluator has not been built");
    }
    else {
      text.append("\tSymmetrical Uncertainty Ranking Filter");
      if (!m_missing_merge) {
        text.append("\n\tMissing values treated as seperate");
      }
    }

    text.append("\n");
    return  text.toString();
  }


  // ============
  // Test method.
  // ============
  /**
   * Main method for testing this class.
   *
   * @param argv should contain the following arguments:
   * -t training file
   */
  public static void main (String[] argv) {
    runEvaluator(new SymmetricalUncertAttributeSetEval(), argv);
  }
}