checkclassifier.java

MacroWeka扩展了著名数据挖掘工具weka

/*
 *    This program is free software; you can redistribute it and/or modify
 *    it under the terms of the GNU General Public License as published by
 *    the Free Software Foundation; either version 2 of the License, or
 *    (at your option) any later version.
 *
 *    This program is distributed in the hope that it will be useful,
 *    but WITHOUT ANY WARRANTY; without even the implied warranty of
 *    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 *    GNU General Public License for more details.
 *
 *    You should have received a copy of the GNU General Public License
 *    along with this program; if not, write to the Free Software
 *    Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */

/*
 *    CheckClassifier.java
 *    Copyright (C) 1999 Len Trigg
 *
 */

package weka.classifiers;

import weka.core.Attribute;
import weka.core.FastVector;
import weka.core.Instance;
import weka.core.Instances;
import weka.core.Option;
import weka.core.OptionHandler;
import weka.core.Utils;
import weka.core.WeightedInstancesHandler;

import java.util.Enumeration;
import java.util.Random;
import java.util.Vector;

/**
 * Class for examining the capabilities and finding problems with 
 * classifiers. If you implement a classifier using the WEKA.libraries,
 * you should run the checks on it to ensure robustness and correct
 * operation. Passing all the tests of this object does not mean
 * bugs in the classifier don't exist, but this will help find some
 * common ones. <p/>
 * 
 * Typical usage: <p/>
 * <code>java weka.classifiers.CheckClassifier -W classifier_name 
 * classifier_options </code><p/>
 * 
 * CheckClassifier reports on the following:
 * <ul>
 *    <li> Classifier abilities <ul>
 *         <li> Possible command line options to the classifier
 *         <li> Whether the classifier can predict nominal and/or predict 
 *              numeric class attributes. Warnings will be displayed if 
 *              performance is worse than ZeroR
 *         <li> Whether the classifier can be trained incrementally
 *         <li> Whether the classifier can handle numeric predictor attributes
 *         <li> Whether the classifier can handle nominal predictor attributes
 *         <li> Whether the classifier can handle string predictor attributes
 *         <li> Whether the classifier can handle missing predictor values
 *         <li> Whether the classifier can handle missing class values
 *         <li> Whether a nominal classifier only handles 2 class problems
 *         <li> Whether the classifier can handle instance weights
 *         </ul>
 *    <li> Correct functioning <ul>
 *         <li> Correct initialisation during buildClassifier (i.e. no result
 *              changes when buildClassifier called repeatedly)
 *         <li> Whether incremental training produces the same results
 *              as during non-incremental training (which may or may not 
 *              be OK)
 *         <li> Whether the classifier alters the data pased to it 
 *              (number of instances, instance order, instance weights, etc)
 *         </ul>
 *    <li> Degenerate cases <ul>
 *         <li> building classifier with zero training instances
 *         <li> all but one predictor attribute values missing
 *         <li> all predictor attribute values missing
 *         <li> all but one class values missing
 *         <li> all class values missing
 *         </ul>
 *    </ul>
 * Running CheckClassifier with the debug option set will output the 
 * training and test datasets for any failed tests.<p/>
 *
 * The <code>weka.classifiers.AbstractClassifierTest</code> uses this
 * class to test all the classifiers. Any changes here, have to be 
 * checked in that abstract test class, too. <p/>
 *
 * Valid options are:<p/>
 *
 * -D <br/>
 * Turn on debugging output.<p/>
 *
 * -S <br/>
 * Silent mode, i.e., no output at all.<p/>
 *
 * -N num <br/>
 * Number of instances to use for datasets (default 20).<p/>
 *
 * -W classname <br/>
 * Specify the full class name of a classifier to perform the 
 * tests on (required).<p/>
 *
 * Options after -- are passed to the designated classifier.<p/>
 *
 * @author Len Trigg ([e-mail])
 * @version $Revision: 1.1 $
 */
public class CheckClassifier implements OptionHandler {

  /*
   * Note about test methods:
   * - return array of booleans
   * - first index: success or not
   * - second index: acceptable or not (e.g., Exception is OK)
   *
   * FracPete (fracpete at waikato dot ac dot nz)
   */
  
  /** 
   * a class for postprocessing the test-data 
   * @see #makeTestDataset(int, int, int, int, int, int, int, int, int, int, boolean)
   */
  public static class PostProcessor {
    /**
     * Provides a hook for derived classes to further modify the data. Currently,
     * the data is just passed through.
     * 
     * @param data	the data to process
     * @return		the processed data
     */
    public Instances process(Instances data) {
      return data;
    }
  }

  /*** The classifier to be examined */
  protected Classifier m_Classifier = new weka.classifiers.rules.ZeroR();

  /** The options to be passed to the base classifier. */
  protected String [] m_ClassifierOptions;

  /** The results of the analysis as a string */
  protected String m_AnalysisResults;

  /** Debugging mode, gives extra output if true */
  protected boolean m_Debug = false;

  /** Silent mode, for no output at all to stdout */
  protected boolean m_Silent = false;

  /** The number of instances in the datasets */
  protected int m_NumInstances = 20;
  
  /** for post-processing the data even further */
  protected PostProcessor m_PostProcessor = null;
  
  /**
   * Returns an enumeration describing the available options.
   *
   * @return an enumeration of all the available options.
   */
  public Enumeration listOptions() {

    Vector newVector = new Vector(2);

    newVector.addElement(new Option(
              "\tTurn on debugging output.",
              "D", 0, "-D"));

    newVector.addElement(new Option(
              "\tSilent mode - prints nothing to stdout.",
              "S", 0, "-S"));

    newVector.addElement(new Option(
	      "\tThe number of instances in the datasets (default 20).",
	      "N", 1, "-N <num>"));

    newVector.addElement(new Option(
	      "\tFull name of the classifier analysed.\n"
	      +"\teg: weka.classifiers.bayes.NaiveBayes",
	      "W", 1, "-W"));

    if ((m_Classifier != null) 
	&& (m_Classifier instanceof OptionHandler)) {
      newVector.addElement(new Option("", "", 0, 
				      "\nOptions specific to classifier "
				      + m_Classifier.getClass().getName()
				      + ":"));
      Enumeration enu = ((OptionHandler)m_Classifier).listOptions();
      while (enu.hasMoreElements())
	newVector.addElement(enu.nextElement());
    }
    
    return newVector.elements();
  }

  /**
   * Parses a given list of options. 
   *
   * @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      tmpStr;
    
    setDebug(Utils.getFlag('D', options));
    
    setSilent(Utils.getFlag('S', options));

    tmpStr = Utils.getOption('N', options);
    if (tmpStr.length() != 0)
      setNumInstances(Integer.parseInt(tmpStr));
    else
      setNumInstances(20);
    
    tmpStr = Utils.getOption('W', options);
    if (tmpStr.length() == 0)
      throw new Exception("A classifier must be specified with"
			  + " the -W option.");
    setClassifier(Classifier.forName(tmpStr, Utils.partitionOptions(options)));
  }

  /**
   * Gets the current settings of the CheckClassifier.
   *
   * @return an array of strings suitable for passing to setOptions
   */
  public String[] getOptions() {
    Vector        result;
    String[]      options;
    int           i;
    
    result = new Vector();

    if (getDebug())
      result.add("-D");

    if (getSilent())
      result.add("-S");

    result.add("-N");
    result.add("" + getNumInstances());
    
    if (getClassifier() != null) {
      result.add("-W");
      result.add(getClassifier().getClass().getName());
    }
    
    if ((m_Classifier != null) && (m_Classifier instanceof OptionHandler))
      options = ((OptionHandler) m_Classifier).getOptions();
    else
      options = new String[0];

    if (options.length > 0) {
      result.add("--");
      for (i = 0; i < options.length; i++)
        result.add(options[i]);
    }
    
    return (String[]) result.toArray(new String[result.size()]);
  }
  
  /**
   * sets the PostProcessor to use
   * 
   * @param value	the new PostProcessor
   * @see #m_PostProcessor
   */
  public void setPostProcessor(PostProcessor value) {
    m_PostProcessor = value;
  }
  
  /**
   * returns the current PostProcessor, can be null
   * 
   * @return		the current PostProcessor
   */
  public PostProcessor getPostProcessor() {
    return m_PostProcessor;
  }

  /**
   * Begin the tests, reporting results to System.out
   */
  public void doTests() {

    if (getClassifier() == null) {
      println("\n=== No classifier set ===");
      return;
    }
    println("\n=== Check on Classifier: "
		       + getClassifier().getClass().getName()
		       + " ===\n");

    // Start tests
    canTakeOptions();
    boolean updateableClassifier = updateableClassifier()[0];
    boolean weightedInstancesHandler = weightedInstancesHandler()[0];
    testsPerClassType(false, updateableClassifier, weightedInstancesHandler);
    testsPerClassType(true, updateableClassifier, weightedInstancesHandler);

  }

  /**
   * Set debugging mode
   *
   * @param debug true if debug output should be printed
   */
  public void setDebug(boolean debug) {
    m_Debug = debug;
    // disable silent mode, if necessary
    if (getDebug())
      setSilent(false);
  }

  /**
   * Get whether debugging is turned on
   *
   * @return true if debugging output is on
   */
  public boolean getDebug() {
    return m_Debug;
  }

  /**
   * Set slient mode, i.e., no output at all to stdout
   *
   * @param value whether silent mode is active or not
   */
  public void setSilent(boolean value) {
    m_Silent = value;
  }

  /**
   * Get whether silent mode is turned on
   *
   * @return true if silent mode is on
   */
  public boolean getSilent() {
    return m_Silent;
  }

  /**
   * Sets the number of instances to use in the datasets (some classifiers
   * might require more instances).
   *
   * @param value the number of instances to use
   */
  public void setNumInstances(int value) {
    m_NumInstances = value;
  }

  /**
   * Gets the current number of instances to use for the datasets.
   *
   * @return the number of instances
   */
  public int getNumInstances() {
    return m_NumInstances;
  }

  /**
   * Set the classifier for boosting. 
   *
   * @param newClassifier the Classifier to use.
   */
  public void setClassifier(Classifier newClassifier) {
    m_Classifier = newClassifier;
  }

  /**
   * Get the classifier used as the classifier
   *
   * @return the classifier used as the classifier
   */
  public Classifier getClassifier() {
    return m_Classifier;
  }

  /**
   * prints the given message to stdout, if not silent mode
   * 
   * @param msg         the text to print to stdout
   */
  protected void print(Object msg) {
    if (!getSilent())
      System.out.print(msg);
  }
  
  /**
   * prints the given message (+ LF) to stdout, if not silent mode
   * 
   * @param msg         the message to println to stdout
   */
  protected void println(Object msg) {
    print(msg + "\n");
  }
  
  /**
   * prints a LF to stdout, if not silent mode
   */
  protected void println() {
    print("\n");
  }
  
  /**
   * Run a battery of tests for a given class attribute type
   *
   * @param numericClass true if the class attribute should be numeric
   * @param updateable true if the classifier is updateable
   * @param weighted true if the classifier says it handles weights
   */
  protected void testsPerClassType(boolean numericClass, boolean updateable,
				   boolean weighted) {

    boolean PNom = canPredict(true, false, false, numericClass)[0];
    boolean PNum = canPredict(false, true, false, numericClass)[0];
    boolean PStr = canPredict(false, false, true, numericClass)[0];
    if (PNom || PNum || PStr) {
      if (weighted)
	instanceWeights(PNom, PNum, PStr, numericClass);

      if (!numericClass)
	canHandleNClasses(PNom, PNum, PStr, 4);

      canHandleZeroTraining(PNom, PNum, PStr, numericClass);
      boolean handleMissingPredictors = canHandleMissing(PNom, PNum, PStr, 
							 numericClass, 
							 true, false, 20)[0];