experiment.java

Java 编写的多种数据挖掘算法 包括聚类、分类、预处理等

    determineAdditionalResultMeasures();
      //    }

    m_ResultProducer.setResultListener(m_ResultListener);
    m_ResultProducer.setAdditionalMeasures(m_AdditionalMeasures);
    m_ResultProducer.preProcess();

    // constrain the additional measures to be only those allowable
    // by the ResultListener
    String [] columnConstraints = m_ResultListener.
      determineColumnConstraints(m_ResultProducer);

    if (columnConstraints != null) {
      m_ResultProducer.setAdditionalMeasures(columnConstraints);
    }
  }

  /**
   * Iterate over the objects in the property array to determine what
   * (if any) additional measures they support
   * 
   * @throws Exception 	if additional measures don't comply to the naming 
   * 			convention (starting with "measure")
   */
  private void determineAdditionalResultMeasures() throws Exception {
    m_AdditionalMeasures = null;
    FastVector measureNames = new FastVector();

    // first try the result producer, then property array if applicable
    if (m_ResultProducer instanceof AdditionalMeasureProducer) {
      Enumeration am = ((AdditionalMeasureProducer)m_ResultProducer).
	enumerateMeasures();
      while (am.hasMoreElements()) {
	String mname = (String)am.nextElement();
	if (mname.startsWith("measure")) {
	  if (measureNames.indexOf(mname) == -1) {
	    measureNames.addElement(mname);
	  }
	} else {
	  throw new Exception ("Additional measures in "
			       + m_ResultProducer.getClass().getName()
			       +" must obey the naming convention"
			       +" of starting with \"measure\"");
	}
      }
    }

    if (m_UsePropertyIterator && (m_PropertyArray != null)) {
      for (int i = 0; i < Array.getLength(m_PropertyArray); i++) {
	Object current = Array.get(m_PropertyArray, i);

	if (current instanceof AdditionalMeasureProducer) {
	  Enumeration am = ((AdditionalMeasureProducer)current).
	    enumerateMeasures();
	  while (am.hasMoreElements()) {
	    String mname = (String)am.nextElement();
	    if (mname.startsWith("measure")) {
	      if (measureNames.indexOf(mname) == -1) {
		measureNames.addElement(mname);
	      }
	    } else {
	      throw new Exception ("Additional measures in "
				   + current.getClass().getName()
				   +" must obey the naming convention"
				   +" of starting with \"measure\"");
	    }
	  }
	}
      }
    }
    if (measureNames.size() > 0) {
      m_AdditionalMeasures = new String [measureNames.size()];
      for (int i=0;i<measureNames.size();i++) {
	m_AdditionalMeasures[i] = (String)measureNames.elementAt(i);
      }
    }
  }

  
  /**
   * Recursively sets the custom property value, by setting all values
   * along the property path.
   *
   * @param propertyDepth the current position along the property path
   * @param origValue the value to set the property to
   * @throws Exception if an error occurs
   */
  protected void setProperty(int propertyDepth, Object origValue)
    throws Exception {
    
    PropertyDescriptor current = m_PropertyPath[propertyDepth].property;
    Object subVal = null;
    if (propertyDepth < m_PropertyPath.length - 1) {
      Method getter = current.getReadMethod();
      Object getArgs [] = { };
      subVal = getter.invoke(origValue, getArgs);
      setProperty(propertyDepth + 1, subVal);
    } else {
      subVal = Array.get(m_PropertyArray, m_PropertyNumber);
    }
    Method setter = current.getWriteMethod();
    Object [] args = { subVal };
    setter.invoke(origValue, args);
  }

  /**
   * Returns true if there are more iterations to carry out in the experiment.
   *
   * @return true if so
   */
  public boolean hasMoreIterations() {

    return !m_Finished;
  }
  
  /**
   * Carries out the next iteration of the experiment.
   *
   * @throws Exception if an error occurs
   */
  public void nextIteration() throws Exception {
    
    if (m_UsePropertyIterator) {
      if (m_CurrentProperty != m_PropertyNumber) {
	setProperty(0, m_ResultProducer);
	m_CurrentProperty = m_PropertyNumber;
      }
    }
    
    if (m_CurrentInstances == null) {
      File currentFile = (File) getDatasets().elementAt(m_DatasetNumber);
      Reader reader = new FileReader(currentFile);
      Instances data = new Instances(new BufferedReader(reader));
      if (m_ClassFirst) {
	data.setClassIndex(0);
      } else {
	data.setClassIndex(data.numAttributes() - 1);
      }
      m_CurrentInstances = data;
      m_ResultProducer.setInstances(m_CurrentInstances);
    }
    
    m_ResultProducer.doRun(m_RunNumber);

    advanceCounters();
  }

  /**
   * Increments iteration counters appropriately.
   */
  public void advanceCounters() {

    if (m_AdvanceDataSetFirst) {
      m_RunNumber ++;
      if (m_RunNumber > getRunUpper()) {
	m_RunNumber = getRunLower();
	m_DatasetNumber ++;
	m_CurrentInstances = null;
	if (m_DatasetNumber >= getDatasets().size()) {
	  m_DatasetNumber = 0;
	  if (m_UsePropertyIterator) {
	    m_PropertyNumber ++;
	    if (m_PropertyNumber >= Array.getLength(m_PropertyArray)) {
	      m_Finished = true;
	    }
	  } else {
	    m_Finished = true;
	  }
	}
      }
    } else { // advance by custom iterator before data set
      m_RunNumber ++;
      if (m_RunNumber > getRunUpper()) {
	m_RunNumber = getRunLower();
	if (m_UsePropertyIterator) {
	  m_PropertyNumber ++;
	  if (m_PropertyNumber >= Array.getLength(m_PropertyArray)) {
	    m_PropertyNumber = 0;
	    m_DatasetNumber ++;
	    m_CurrentInstances = null;
	    if (m_DatasetNumber >= getDatasets().size()) {
	      m_Finished = true;
	    } 
	  }
	} else {
	  m_DatasetNumber ++;
	  m_CurrentInstances = null;
	  if (m_DatasetNumber >= getDatasets().size()) {
	    m_Finished = true;
	  }
	}
      }
    }
  }

  /**
   * Runs all iterations of the experiment, continuing past errors.
   */
  public void runExperiment() {

    while (hasMoreIterations()) {
      try {
	nextIteration();
      } catch (Exception ex) {
	ex.printStackTrace();
	System.err.println(ex.getMessage());
	advanceCounters(); // Try to keep plowing through
      }
    }
  }

  /**
   * Signals that the experiment is finished running, so that cleanup
   * can be done.
   *
   * @throws Exception if an error occurs
   */
  public void postProcess() throws Exception {

    m_ResultProducer.postProcess();
  }
  
  /**
   * Gets the datasets in the experiment.
   *
   * @return the datasets in the experiment.
   */
  public DefaultListModel getDatasets() {
    return m_Datasets;
  }

  /**
   * Set the datasets to use in the experiment
   * @param ds the list of datasets to use
   */
  public void setDatasets(DefaultListModel ds) {
    m_Datasets = ds;
  }

  /**
   * Gets the result listener where results will be sent.
   *
   * @return the result listener where results will be sent.
   */
  public ResultListener getResultListener() {
    
    return m_ResultListener;
  }
  
  /**
   * Sets the result listener where results will be sent.
   *
   * @param newResultListener the result listener where results will be sent.
   */
  public void setResultListener(ResultListener newResultListener) {
    
    m_ResultListener = newResultListener;
  }
  
  /**
   * Get the result producer used for the current experiment.
   *
   * @return the result producer used for the current experiment.
   */
  public ResultProducer getResultProducer() {
    
    return m_ResultProducer;
  }
  
  /**
   * Set the result producer used for the current experiment.
   *
   * @param newResultProducer result producer to use for the current 
   * experiment.
   */
  public void setResultProducer(ResultProducer newResultProducer) {
    
    m_ResultProducer = newResultProducer;
  }
  
  /**
   * Get the upper run number for the experiment.
   *
   * @return the upper run number for the experiment.
   */
  public int getRunUpper() {
    
    return m_RunUpper;
  }
  
  /**
   * Set the upper run number for the experiment.
   *
   * @param newRunUpper the upper run number for the experiment.
   */
  public void setRunUpper(int newRunUpper) {
    
    m_RunUpper = newRunUpper;
  }
  
  /**
   * Get the lower run number for the experiment.
   *
   * @return the lower run number for the experiment.
   */
  public int getRunLower() {
    
    return m_RunLower;
  }
  
  /**
   * Set the lower run number for the experiment.
   *
   * @param newRunLower the lower run number for the experiment.
   */
  public void setRunLower(int newRunLower) {
    
    m_RunLower = newRunLower;
  }

  
  /**
   * Get the user notes.
   *
   * @return User notes associated with the experiment.
   */
  public String getNotes() {
    
    return m_Notes;
  }
  
  /**
   * Set the user notes.
   *
   * @param newNotes New user notes.
   */
  public void setNotes(String newNotes) {
    
    m_Notes = newNotes;
  }
  
  /**
   * Returns an enumeration describing the available options..
   *
   * @return an enumeration of all the available options.
   */
  public Enumeration listOptions() {

    Vector newVector = new Vector(6);

    newVector.addElement(new Option(
	     "\tThe lower run number to start the experiment from.\n"
	      +"\t(default 1)", 
	     "L", 1, 
	     "-L <num>"));
    newVector.addElement(new Option(
	     "\tThe upper run number to end the experiment at (inclusive).\n"
	      +"\t(default 10)", 
	     "U", 1, 
	     "-U <num>"));
    newVector.addElement(new Option(
	     "\tThe dataset to run the experiment on.\n"
	     + "\t(required, may be specified multiple times)", 
	     "T", 1, 
	     "-T <arff file>"));
    newVector.addElement(new Option(
	     "\tThe full class name of a ResultProducer (required).\n"
	      +"\teg: weka.experiment.RandomSplitResultProducer", 
	     "P", 1, 
	     "-P <class name>"));
    newVector.addElement(new Option(
	     "\tThe full class name of a ResultListener (required).\n"
	      +"\teg: weka.experiment.CSVResultListener", 
	     "D", 1, 
	     "-D <class name>"));
    newVector.addElement(new Option(
	     "\tA string containing any notes about the experiment.\n"
	      +"\t(default none)", 
	     "N", 1, 
	     "-N <string>"));

    if ((m_ResultProducer != null) &&
	(m_ResultProducer instanceof OptionHandler)) {
      newVector.addElement(new Option(
	     "",
	     "", 0, "\nOptions specific to result producer "
	     + m_ResultProducer.getClass().getName() + ":"));
      Enumeration enm = ((OptionHandler)m_ResultProducer).listOptions();
      while (enm.hasMoreElements()) {
	newVector.addElement(enm.nextElement());
      }