discretizefilter.java

一个数据挖掘系统的源码

   * displaying in the explorer/experimenter gui
   */
  public String makeBinaryTipText() {

    return "Make resulting attributes binary.";
  }

  /**
   * Gets whether binary attributes should be made for discretized ones.
   *
   * @return true if attributes will be binarized
   */
  public boolean getMakeBinary() {

    return m_MakeBinary;
  }

  /**
   * Sets whether binary attributes should be made for discretized ones.
   *
   * @param makeBinary if binary attributes are to be made
   */
  public void setMakeBinary(boolean makeBinary) {

    m_MakeBinary = makeBinary;
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String useMDLTipText() {

    return "Use class-based discretization. If set to false, does"
      + " not require a class attribute, and uses a fixed number"
      + " of bins (according to bins setting).";
  }

  /**
   * Gets whether MDL will be used as the discretisation method.
   *
   * @return true if so, false if fixed bins should be used.
   */
  public boolean getUseMDL() {

    return m_UseMDL;
  }

  /**
   * Sets whether MDL will be used as the discretisation method.
   *
   * @param useMDL true if MDL should be used, false if fixed bins should
   * be used.
   */
  public void setUseMDL(boolean useMDL) {

    m_UseMDL = useMDL;
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String useKononenkoTipText() {

    return "Use Kononenko's MDL criterion. If set to false"
      + " uses the Fayyad & Irani criterion.";
  }

  /**
   * Get the value of UseEqualFrequency.
   *
   * @return Value of UseEqualFrequency.
   */
  public boolean getUseEqualFrequency() {

    return m_UseEqualFrequency;
  }

  /**
   * Set the value of UseEqualFrequency.
   *
   * @param newUseEqualFrequency Value to assign to UseEqualFrequency.
   */
  public void setUseEqualFrequency(boolean newUseEqualFrequency) {

    m_UseEqualFrequency = newUseEqualFrequency;
  }

  /**
   * Gets whether Kononenko's MDL criterion is to be used.
   *
   * @return true if Kononenko's criterion will be used.
   */
  public boolean getUseKononenko() {

    return m_UseKononenko;
  }

  /**
   * Sets whether Kononenko's MDL criterion is to be used.
   *
   * @param useKon true if Kononenko's one is to be used
   */
  public void setUseKononenko(boolean useKon) {

    m_UseKononenko = useKon;
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String useBetterEncodingTipText() {

    return "Uses a more efficient split point encoding.";
  }

  /**
   * Gets whether better encoding is to be used for MDL.
   *
   * @return true if the better MDL encoding will be used
   */
  public boolean getUseBetterEncoding() {

    return m_UseBetterEncoding;
  }

  /**
   * Sets whether better encoding is to be used for MDL.
   *
   * @param useBetterEncoding true if better encoding to be used.
   */
  public void setUseBetterEncoding(boolean useBetterEncoding) {

    m_UseBetterEncoding = useBetterEncoding;
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String binsTipText() {

    return "Number of bins for class-blind discretisation. This"
      + " setting is ignored if MDL-based discretisation is used.";
  }

  /**
   * Gets the number of bins numeric attributes will be divided into
   *
   * @return the number of bins.
   */
  public int getBins() {

    return m_NumBins;
  }

  /**
   * Sets the number of bins to divide each selected numeric attribute into
   *
   * @param numBins the number of bins
   */
  public void setBins(int numBins) {

    m_NumBins = numBins;
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String invertSelectionTipText() {

    return "Set attribute selection mode. If false, only selected"
      + " (numeric) attributes in the range will be discretized; if"
      + " true, only non-selected attributes will be discretized.";
  }

  /**
   * Gets whether the supplied columns are to be removed or kept
   *
   * @return true if the supplied columns will be kept
   */
  public boolean getInvertSelection() {

    return m_DiscretizeCols.getInvert();
  }

  /**
   * Sets whether selected columns should be removed or kept. If true the
   * selected columns are kept and unselected columns are deleted. If false
   * selected columns are deleted and unselected columns are kept.
   *
   * @param invert the new invert setting
   */
  public void setInvertSelection(boolean invert) {

    m_DiscretizeCols.setInvert(invert);
  }

  /**
   * Returns the tip text for this property
   *
   * @return tip text for this property suitable for
   * displaying in the explorer/experimenter gui
   */
  public String attributeIndicesTipText() {
    return "Specify range of attributes to act on."
      + " This is a comma separated list of attribute indices, with"
      + " \"first\" and \"last\" valid values. Specify an inclusive"
      + " range with \"-\". E.g: \"first-3,5,6-10,last\".";
  }

  /**
   * Gets the current range selection
   *
   * @return a string containing a comma separated list of ranges
   */
  public String getAttributeIndices() {

    return m_DiscretizeCols.getRanges();
  }

  /**
   * Sets which attributes are to be Discretized (only numeric
   * attributes among the selection will be Discretized).
   *
   * @param rangeList a string representing the list of attributes. Since
   * the string will typically come from a user, attributes are indexed from
   * 1. <br>
   * eg: first-3,5,6-last
   * @exception IllegalArgumentException if an invalid range list is supplied
   */
  public void setAttributeIndices(String rangeList) {

    m_DiscretizeCols.setRanges(rangeList);
  }

  /**
   * Sets which attributes are to be Discretized (only numeric
   * attributes among the selection will be Discretized).
   *
   * @param attributes an array containing indexes of attributes to Discretize.
   * Since the array will typically come from a program, attributes are indexed
   * from 0.
   * @exception IllegalArgumentException if an invalid set of ranges
   * is supplied
   */
  public void setAttributeIndicesArray(int [] attributes) {

    setAttributeIndices(Range.indicesToRangeList(attributes));
  }

  /**
   * Gets the cut points for an attribute
   *
   * @param the index (from 0) of the attribute to get the cut points of
   * @return an array containing the cutpoints (or null if the
   * attribute requested isn't being Discretized
   */
  public double [] getCutPoints(int attributeIndex) {

    if (m_CutPoints == null) {
      return null;
    }
    return m_CutPoints[attributeIndex];
  }

  /** Generate the cutpoints for each attribute */
  protected void calculateCutPoints() throws Exception{

    Instances copy = null;

    m_CutPoints = new double [getInputFormat().numAttributes()] [];
    for(int i = getInputFormat().numAttributes() - 1; i >= 0; i--) {
      if ((m_DiscretizeCols.isInRange(i)) &&
	  (getInputFormat().attribute(i).isNumeric())) {
	if (m_UseMDL) {

	  // Use copy to preserve order
	  if (copy == null) {
	    copy = new Instances(getInputFormat());
	  }
	  calculateCutPointsByMDL(i, copy);
	} else {
	  if (m_FindNumBins) {
	    findNumBins(i);
	  } else if (!m_UseEqualFrequency) {
	    calculateCutPointsByEqualWidthBinning(i);
	  } else {
	    calculateCutPointsByEqualFrequencyBinning(i);
	  }
	}
      }
    }
  }

  /**
   * Set cutpoints for a single attribute using MDL.
   *
   * @param index the index of the attribute to set cutpoints for
   */
  protected void calculateCutPointsByMDL(int index,
					 Instances data) throws Exception {

    // Sort instances
    data.sort(data.attribute(index));

    // Find first instances that's missing
    int firstMissing = data.numInstances();
    for (int i = 0; i < data.numInstances(); i++) {
      if (data.instance(i).isMissing(index)) {
        firstMissing = i;
        break;
      }
    }
    m_CutPoints[index] = cutPointsForSubset(data, index, 0, firstMissing);
  }

  /** Test using Kononenko's MDL criterion. */
  private boolean KononenkosMDL(double[] priorCounts,
				double[][] bestCounts,
				double numInstances,
				int numCutPoints) {

    double distPrior, instPrior, distAfter = 0, sum, instAfter = 0;
    double before, after;
    int numClassesTotal;

    // Number of classes occuring in the set
    numClassesTotal = 0;
    for (int i = 0; i < priorCounts.length; i++) {
      if (priorCounts[i] > 0) {
	numClassesTotal++;
      }
    }

    // Encode distribution prior to split
    distPrior = SpecialFunctions.log2Binomial(numInstances
					      + numClassesTotal - 1,
					      numClassesTotal - 1);

    // Encode instances prior to split.
    instPrior = SpecialFunctions.log2Multinomial(numInstances,
						 priorCounts);

    before = instPrior + distPrior;

    // Encode distributions and instances after split.
    for (int i = 0; i < bestCounts.length; i++) {
      sum = Utils.sum(bestCounts[i]);
      distAfter += SpecialFunctions.log2Binomial(sum + numClassesTotal - 1,
						 numClassesTotal - 1);
      instAfter += SpecialFunctions.log2Multinomial(sum,
						    bestCounts[i]);
    }

    // Coding cost after split
    after = Utils.log2(numCutPoints) + distAfter + instAfter;

    // Check if split is to be accepted
    return (Utils.gr(before, after));
  }


  /** Test using Fayyad and Irani's MDL criterion. */
  private boolean FayyadAndIranisMDL(double[] priorCounts,
				     double[][] bestCounts,
				     double numInstances,
				     int numCutPoints) {

    double priorEntropy, entropy, gain;
    double entropyLeft, entropyRight, delta;
    int numClassesTotal, numClassesRight, numClassesLeft;

    // Compute entropy before split.
    priorEntropy = ContingencyTables.entropy(priorCounts);

    // Compute entropy after split.
    entropy = ContingencyTables.entropyConditionedOnRows(bestCounts);

    // Compute information gain.
    gain = priorEntropy - entropy;

    // Number of classes occuring in the set
    numClassesTotal = 0;
    for (int i = 0; i < priorCounts.length; i++) {
      if (priorCounts[i] > 0) {
	numClassesTotal++;
      }
    }

    // Number of classes occuring in the left subset
    numClassesLeft = 0;
    for (int i = 0; i < bestCounts[0].length; i++) {
      if (bestCounts[0][i] > 0) {
	numClassesLeft++;
      }
    }

    // Number of classes occuring in the right subset
    numClassesRight = 0;
    for (int i = 0; i < bestCounts[1].length; i++) {
      if (bestCounts[1][i] > 0) {
	numClassesRight++;
      }
    }

    // Entropy of the left and the right subsets
    entropyLeft = ContingencyTables.entropy(bestCounts[0]);
    entropyRight = ContingencyTables.entropy(bestCounts[1]);

    // Compute terms for MDL formula
    delta = Utils.log2(Math.pow(3, numClassesTotal) - 2) -
      (((double) numClassesTotal * priorEntropy) -
       (numClassesRight * entropyRight) -
       (numClassesLeft * entropyLeft));

    // Check if split is to be accepted
    return (Utils.gr(gain, (Utils.log2(numCutPoints) + delta) /
		     (double)numInstances));
  }


  /** Selects cutpoints for sorted subset. */
  private double[] cutPointsForSubset(Instances instances, int attIndex,