greedystepwise.java

这是关于数据挖掘的一些算法

   *  Specify a starting set of attributes.
   *  Eg. 1,3,5-7.</pre>
   * 
   * <pre> -R
   *  Produce a ranked list of attributes.</pre>
   * 
   * <pre> -T &lt;threshold&gt;
   *  Specify a theshold by which attributes
   *  may be discarded from the ranking.
   *  Use in conjuction with -R</pre>
   * 
   * <pre> -N &lt;num to select&gt;
   *  Specify number of attributes to select</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 {
    String optionString;
    resetOptions();

    setSearchBackwards(Utils.getFlag('B', options));

    setConservativeForwardSelection(Utils.getFlag('C', options));

    optionString = Utils.getOption('P', options);
    if (optionString.length() != 0) {
      setStartSet(optionString);
    }

    setGenerateRanking(Utils.getFlag('R', options));

    optionString = Utils.getOption('T', options);
    if (optionString.length() != 0) {
      Double temp;
      temp = Double.valueOf(optionString);
      setThreshold(temp.doubleValue());
    }

    optionString = Utils.getOption('N', options);
    if (optionString.length() != 0) {
      setNumToSelect(Integer.parseInt(optionString));
    }
  }

  /**
   * Gets the current settings of ReliefFAttributeEval.
   *
   * @return an array of strings suitable for passing to setOptions()
   */
  public String[] getOptions () {
    String[] options = new String[9];
    int current = 0;
    
    if (getSearchBackwards()) {
      options[current++] = "-B";
    }

    if (getConservativeForwardSelection()) {
      options[current++] = "-C";
    }

    if (!(getStartSet().equals(""))) {
      options[current++] = "-P";
      options[current++] = ""+startSetToString();
    }

    if (getGenerateRanking()) {
      options[current++] = "-R";
    }
    options[current++] = "-T";
    options[current++] = "" + getThreshold();

    options[current++] = "-N";
    options[current++] = ""+getNumToSelect();

    while (current < options.length) {
      options[current++] = "";
    }
    return  options;
  }

  /**
   * converts the array of starting attributes to a string. This is
   * used by getOptions to return the actual attributes specified
   * as the starting set. This is better than using m_startRanges.getRanges()
   * as the same start set can be specified in different ways from the
   * command line---eg 1,2,3 == 1-3. This is to ensure that stuff that
   * is stored in a database is comparable.
   * @return a comma seperated list of individual attribute numbers as a String
   */
  protected String startSetToString() {
    StringBuffer FString = new StringBuffer();
    boolean didPrint;
    
    if (m_starting == null) {
      return getStartSet();
    }
    for (int i = 0; i < m_starting.length; i++) {
      didPrint = false;
      
      if ((m_hasClass == false) || 
	  (m_hasClass == true && i != m_classIndex)) {
	FString.append((m_starting[i] + 1));
	didPrint = true;
      }
      
      if (i == (m_starting.length - 1)) {
	FString.append("");
      }
      else {
	if (didPrint) {
	  FString.append(",");
	  }
      }
    }

    return FString.toString();
  }

  /**
   * returns a description of the search.
   * @return a description of the search as a String.
   */
  public String toString() {
    StringBuffer FString = new StringBuffer();
    FString.append("\tGreedy Stepwise ("
		   + ((m_backward)
		      ? "backwards)"
		      : "forwards)")+".\n\tStart set: ");

    if (m_starting == null) {
      if (m_backward) {
	FString.append("all attributes\n");
      } else {
	FString.append("no attributes\n");
      }
    }
    else {
      FString.append(startSetToString()+"\n");
    }
    if (!m_doneRanking) {
      FString.append("\tMerit of best subset found: "
		     +Utils.doubleToString(Math.abs(m_bestMerit),8,3)+"\n");
    }
    
    if ((m_threshold != -Double.MAX_VALUE) && (m_doneRanking)) {
      FString.append("\tThreshold for discarding attributes: "
		     + Utils.doubleToString(m_threshold,8,4)+"\n");
    }

    return FString.toString();
  }


  /**
   * Searches the attribute subset space by forward selection.
   *
   * @param ASEval the attribute evaluator to guide the search
   * @param data the training instances.
   * @return an array (not necessarily ordered) of selected attribute indexes
   * @throws Exception if the search can't be completed
   */
  public int[] search (ASEvaluation ASEval, Instances data)
    throws Exception {

    int i;
    double best_merit = -Double.MAX_VALUE;
    double temp_best,temp_merit;
    int temp_index=0;
    BitSet temp_group;

    if (data != null) { // this is a fresh run so reset
      resetOptions();
      m_Instances = data;
    }
    m_ASEval = ASEval;

    m_numAttribs = m_Instances.numAttributes();

    if (m_best_group == null) {
      m_best_group = new BitSet(m_numAttribs);
    }

    if (!(m_ASEval instanceof SubsetEvaluator)) {
      throw  new Exception(m_ASEval.getClass().getName() 
			   + " is not a " 
			   + "Subset evaluator!");
    }

    m_startRange.setUpper(m_numAttribs-1);
    if (!(getStartSet().equals(""))) {
      m_starting = m_startRange.getSelection();
    }

    if (m_ASEval instanceof UnsupervisedSubsetEvaluator) {
      m_hasClass = false;
      m_classIndex = -1;
    }
    else {
      m_hasClass = true;
      m_classIndex = m_Instances.classIndex();
    }

    SubsetEvaluator ASEvaluator = (SubsetEvaluator)m_ASEval;

    if (m_rankedAtts == null) {
      m_rankedAtts = new double[m_numAttribs][2];
      m_rankedSoFar = 0;
    }

    // If a starting subset has been supplied, then initialise the bitset
    if (m_starting != null && m_rankedSoFar <= 0) {
      for (i = 0; i < m_starting.length; i++) {
	if ((m_starting[i]) != m_classIndex) {
	  m_best_group.set(m_starting[i]);
	}
      }
    } else {
      if (m_backward && m_rankedSoFar <= 0) {
	for (i = 0; i < m_numAttribs; i++) {
	  if (i != m_classIndex) {
	    m_best_group.set(i);
	  }
	}
      }
    }

    // Evaluate the initial subset
    best_merit = ASEvaluator.evaluateSubset(m_best_group);

    // main search loop
    boolean done = false;
    boolean addone = false;
    boolean z;
    while (!done) {
      temp_group = (BitSet)m_best_group.clone();
      temp_best = best_merit;
      if (m_doRank) {
	temp_best = -Double.MAX_VALUE;
      }
      done = true;
      addone = false;
      for (i=0;i<m_numAttribs;i++) {
	if (m_backward) {
	  z = ((i != m_classIndex) && (temp_group.get(i)));
	} else {
	  z = ((i != m_classIndex) && (!temp_group.get(i)));
	}
	if (z) {
	  // set/unset the bit
	  if (m_backward) {
	    temp_group.clear(i);
	  } else {
	    temp_group.set(i);
	  }
	  temp_merit = ASEvaluator.evaluateSubset(temp_group);
	  if (m_backward) {
	    z = (temp_merit >= temp_best);
	  } else {
            if (m_conservativeSelection) {
              z = (temp_merit >= temp_best);
            } else {
              z = (temp_merit > temp_best);
            }
	  }

	  if (z) {
            temp_best = temp_merit;
            temp_index = i;
            addone = true;
            done = false;
	  }

	  // unset this addition/deletion
	  if (m_backward) {
	    temp_group.set(i);
	  } else {
	    temp_group.clear(i);
	  }
	  if (m_doRank) {
	    done = false;
	  }
	}
      }
      if (addone) {
	if (m_backward) {
	  m_best_group.clear(temp_index);
	} else {
	  m_best_group.set(temp_index);
	}
	best_merit = temp_best;
	m_rankedAtts[m_rankedSoFar][0] = temp_index;
	m_rankedAtts[m_rankedSoFar][1] = best_merit;
	m_rankedSoFar++;
      }
    }
    m_bestMerit = best_merit;
    return attributeList(m_best_group);
  }

  /**
   * Produces a ranked list of attributes. Search must have been performed
   * prior to calling this function. Search is called by this function to
   * complete the traversal of the the search space. A list of
   * attributes and merits are returned. The attributes a ranked by the
   * order they are added to the subset during a forward selection search.
   * Individual merit values reflect the merit associated with adding the
   * corresponding attribute to the subset; because of this, merit values
   * may initially increase but then decrease as the best subset is
   * "passed by" on the way to the far side of the search space.
   *
   * @return an array of attribute indexes and associated merit values
   * @throws Exception if something goes wrong.
   */
  public double [][] rankedAttributes() throws Exception {
    
    if (m_rankedAtts == null || m_rankedSoFar == -1) {
      throw new Exception("Search must be performed before attributes "
			  +"can be ranked.");
    }
    
    m_doRank = true;
    search (m_ASEval, null);

    double [][] final_rank = new double [m_rankedSoFar][2];
    for (int i=0;i<m_rankedSoFar;i++) {
      final_rank[i][0] = m_rankedAtts[i][0];
      final_rank[i][1] = m_rankedAtts[i][1];
    }
    
    resetOptions();
    m_doneRanking = true;

    if (m_numToSelect > final_rank.length) {
      throw new Exception("More attributes requested than exist in the data");
    }

    if (m_numToSelect <= 0) {
      if (m_threshold == -Double.MAX_VALUE) {
	m_calculatedNumToSelect = final_rank.length;
      } else {
	determineNumToSelectFromThreshold(final_rank);
      }
    }

    return final_rank;
  }

  private void determineNumToSelectFromThreshold(double [][] ranking) {
    int count = 0;
    for (int i = 0; i < ranking.length; i++) {
      if (ranking[i][1] > m_threshold) {
	count++;
      }
    }
    m_calculatedNumToSelect = count;
  }

  /**
   * converts a BitSet into a list of attribute indexes 
   * @param group the BitSet to convert
   * @return an array of attribute indexes
   **/
  protected int[] attributeList (BitSet group) {
    int count = 0;

    // count how many were selected
    for (int i = 0; i < m_numAttribs; i++) {
      if (group.get(i)) {
	count++;
      }
    }

    int[] list = new int[count];
    count = 0;

    for (int i = 0; i < m_numAttribs; i++) {
      if (group.get(i)) {
	list[count++] = i;
      }
    }

    return  list;
  }

  /**
   * Resets options
   */
  protected void resetOptions() {
    m_doRank = false;
    m_best_group = null;
    m_ASEval = null;
    m_Instances = null;
    m_rankedSoFar = -1;
    m_rankedAtts = null;
  }
}