classifierpanel.java

wekaUT是 university texas austin 开发的基于weka的半指导学习(semi supervised learning)的分类器

/*
 *    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.
 */

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


package weka.gui.explorer;

import weka.core.Instances;
import weka.core.Instance;
import weka.core.FastVector;
import weka.core.OptionHandler;
import weka.core.Attribute;
import weka.core.Utils;
import weka.core.Drawable;
import weka.core.SerializedObject;
import weka.classifiers.Classifier;
import weka.classifiers.DistributionClassifier;
import weka.classifiers.Evaluation;
import weka.classifiers.CostMatrix;
import weka.classifiers.evaluation.NominalPrediction;
import weka.classifiers.evaluation.MarginCurve;
import weka.classifiers.evaluation.ThresholdCurve;
import weka.classifiers.evaluation.CostCurve;
import weka.filters.Filter;
import weka.gui.Logger;
import weka.gui.TaskLogger;
import weka.gui.SysErrLog;
import weka.gui.GenericObjectEditor;
import weka.gui.PropertyPanel;
import weka.gui.ResultHistoryPanel;
import weka.gui.SetInstancesPanel;
import weka.gui.CostMatrixEditor;
import weka.gui.PropertyDialog;
import weka.gui.InstancesSummaryPanel;
import weka.gui.SaveBuffer;
import weka.gui.visualize.VisualizePanel;
import weka.gui.visualize.PlotData2D;
import weka.gui.visualize.Plot2D;
import weka.gui.ExtensionFileFilter;

import weka.gui.treevisualizer.*;

import java.util.Random;
import java.util.Date;
import java.text.SimpleDateFormat;
import java.awt.FlowLayout;
import java.awt.BorderLayout;
import java.awt.GridLayout;
import java.awt.GridBagLayout;
import java.awt.GridBagConstraints;
import java.awt.Insets;
import java.awt.Font;
import java.awt.Point;
import java.awt.event.ActionListener;
import java.awt.event.ActionEvent;
import java.awt.event.InputEvent;
import java.awt.event.MouseAdapter;
import java.awt.event.MouseEvent;
import java.awt.Window;
import java.awt.Dimension;
import java.beans.PropertyChangeListener;
import java.beans.PropertyChangeEvent;
import java.beans.PropertyChangeSupport;

import java.io.File;
import java.io.FileWriter;
import java.io.Writer;
import java.io.BufferedWriter;
import java.io.PrintWriter;

import java.io.OutputStream;
import java.io.ObjectOutputStream;
import java.io.FileOutputStream;
import java.util.zip.GZIPOutputStream;
import java.io.InputStream;
import java.io.ObjectInputStream;
import java.io.FileInputStream;
import java.util.zip.GZIPInputStream;

import javax.swing.JFileChooser;
import javax.swing.JPanel;
import javax.swing.JLabel;
import javax.swing.JButton;
import javax.swing.BorderFactory;
import javax.swing.JTextArea;
import javax.swing.JScrollPane;
import javax.swing.JRadioButton;
import javax.swing.ButtonGroup;
import javax.swing.JOptionPane;
import javax.swing.JComboBox;
import javax.swing.DefaultComboBoxModel;
import javax.swing.JTextField;
import javax.swing.SwingConstants;
import javax.swing.JFrame;
import javax.swing.event.ChangeListener;
import javax.swing.event.ChangeEvent;
import javax.swing.JViewport;
import javax.swing.JCheckBox;
import javax.swing.ListSelectionModel;
import javax.swing.event.ListSelectionEvent;
import javax.swing.event.ListSelectionListener;
import javax.swing.JPopupMenu;
import javax.swing.JMenu;
import javax.swing.JMenuItem;
import javax.swing.filechooser.FileFilter;

/** 
 * This panel allows the user to select and configure a classifier, set the
 * attribute of the current dataset to be used as the class, and evaluate
 * the classifier using a number of testing modes (test on the training data,
 * train/test on a percentage split, n-fold cross-validation, test on a
 * separate split). The results of classification runs are stored in a result
 * history so that previous results are accessible.
 *
 * @author Len Trigg ([e-mail])
 * @author Mark Hall ([e-mail])
 * @author Richard Kirkby ([e-mail])
 * @version $Revision: 1.1.1.1 $
 */
public class ClassifierPanel extends JPanel {

  /** Lets the user configure the classifier */
  protected GenericObjectEditor m_ClassifierEditor =
    new GenericObjectEditor();

  /** The panel showing the current classifier selection */
  protected PropertyPanel m_CEPanel = new PropertyPanel(m_ClassifierEditor);
  
  /** The output area for classification results */
  protected JTextArea m_OutText = new JTextArea(20, 40);

  /** The destination for log/status messages */
  protected Logger m_Log = new SysErrLog();

  /** The buffer saving object for saving output */
  SaveBuffer m_SaveOut = new SaveBuffer(m_Log, this);

  /** A panel controlling results viewing */
  protected ResultHistoryPanel m_History = new ResultHistoryPanel(m_OutText);

  /** Lets the user select the class column */
  protected JComboBox m_ClassCombo = new JComboBox();

  /** Click to set test mode to cross-validation */
  protected JRadioButton m_CVBut = new JRadioButton("Cross-validation");

  /** Click to set test mode to generate a % split */
  protected JRadioButton m_PercentBut = new JRadioButton("Percentage split");

  /** Click to set test mode to test on training data */
  protected JRadioButton m_TrainBut = new JRadioButton("Use training set");

  /** Click to set test mode to a user-specified test set */
  protected JRadioButton m_TestSplitBut =
    new JRadioButton("Supplied test set");

  /** Check to save the predictions in the results list for visualizing
      later on */
  protected JCheckBox m_StorePredictionsBut = 
    new JCheckBox("Store predictions for visualization");

  /** Check to output the model built from the training data */
  protected JCheckBox m_OutputModelBut = new JCheckBox("Output model");

  /** Check to output true/false positives, precision/recall for each class */
  protected JCheckBox m_OutputPerClassBut =
    new JCheckBox("Output per-class stats");

  /** Check to output a confusion matrix */
  protected JCheckBox m_OutputConfusionBut =
    new JCheckBox("Output confusion matrix");

  /** Check to output entropy statistics */
  protected JCheckBox m_OutputEntropyBut =
    new JCheckBox("Output entropy evaluation measures");

  /** Check to output text predictions */
  protected JCheckBox m_OutputPredictionsTextBut =
    new JCheckBox("Output text predictions on test set");

  /** Check to evaluate w.r.t a cost matrix */
  protected JCheckBox m_EvalWRTCostsBut =
    new JCheckBox("Cost-sensitive evaluation");

  protected JButton m_SetCostsBut = new JButton("Set...");

  /** Label by where the cv folds are entered */
  protected JLabel m_CVLab = new JLabel("Folds", SwingConstants.RIGHT);

  /** The field where the cv folds are entered */
  protected JTextField m_CVText = new JTextField("10");

  /** Label by where the % split is entered */
  protected JLabel m_PercentLab = new JLabel("%", SwingConstants.RIGHT);

  /** The field where the % split is entered */
  protected JTextField m_PercentText = new JTextField("66");

  /** The button used to open a separate test dataset */
  protected JButton m_SetTestBut = new JButton("Set...");

  /** The frame used to show the test set selection panel */
  protected JFrame m_SetTestFrame;

  /** The frame used to show the cost matrix editing panel */
  protected PropertyDialog m_SetCostsFrame;

  /**
   * Alters the enabled/disabled status of elements associated with each
   * radio button
   */
  ActionListener m_RadioListener = new ActionListener() {
    public void actionPerformed(ActionEvent e) {
      updateRadioLinks();
    }
  };

  /** Button for further output/visualize options */
  JButton m_MoreOptions = new JButton("More options...");

  /**
   * User specified random seed for cross validation or % split
   */
  protected JTextField m_RandomSeedText = new JTextField("1      ");
  protected JLabel m_RandomLab = new JLabel("Random seed for XVal / % Split", 
					    SwingConstants.RIGHT);

  /** Click to start running the classifier */
  protected JButton m_StartBut = new JButton("Start");

  /** Click to stop a running classifier */
  protected JButton m_StopBut = new JButton("Stop");

  /** Stop the class combo from taking up to much space */
  private Dimension COMBO_SIZE = new Dimension(150, m_StartBut
					       .getPreferredSize().height);

  /** The cost matrix editor for evaluation costs */
  protected CostMatrixEditor m_CostMatrixEditor = new CostMatrixEditor();

  /** The main set of instances we're playing with */
  protected Instances m_Instances;

  /** The user-supplied test set (if any) */
  protected Instances m_TestInstances;

  /** The user supplied test set after preprocess filters have been applied */
  protected Instances m_TestInstancesCopy;
  
  /** A thread that classification runs in */
  protected Thread m_RunThread;

  /** default x index for visualizing */
  protected int m_visXIndex;
  
  /** default y index for visualizing */
  protected int m_visYIndex;

  /** The current visualization object */
  protected VisualizePanel m_CurrentVis = null;

  /** The instances summary panel displayed by m_SetTestFrame */
  protected InstancesSummaryPanel m_Summary = null;

  /** Filter to ensure only model files are selected */  
  protected FileFilter m_ModelFilter =
    new ExtensionFileFilter("model", "Model object files");

  /** The file chooser for selecting model files */
  protected JFileChooser m_FileChooser 
    = new JFileChooser(new File(System.getProperty("user.dir")));

  /* Register the property editors we need */
  static {
    java.beans.PropertyEditorManager
      .registerEditor(weka.core.SelectedTag.class,
		      weka.gui.SelectedTagEditor.class);
    java.beans.PropertyEditorManager
      .registerEditor(weka.filters.Filter.class,
		      weka.gui.GenericObjectEditor.class);
    java.beans.PropertyEditorManager
      .registerEditor(weka.classifiers.Classifier [].class,
		      weka.gui.GenericArrayEditor.class);
    java.beans.PropertyEditorManager
      .registerEditor(weka.classifiers.DistributionClassifier.class,
		      weka.gui.GenericObjectEditor.class);
    java.beans.PropertyEditorManager
      .registerEditor(weka.classifiers.Classifier.class,
		      weka.gui.GenericObjectEditor.class);
    java.beans.PropertyEditorManager
      .registerEditor(weka.classifiers.CostMatrix.class,
		      weka.gui.CostMatrixEditor.class);
  }
  
  /**
   * Creates the classifier panel
   */
  public ClassifierPanel() {

    // Connect / configure the components
    m_OutText.setEditable(false);
    m_OutText.setFont(new Font("Monospaced", Font.PLAIN, 12));
    m_OutText.setBorder(BorderFactory.createEmptyBorder(5, 5, 5, 5));
    m_OutText.addMouseListener(new MouseAdapter() {
      public void mouseClicked(MouseEvent e) {
	if ((e.getModifiers() & InputEvent.BUTTON1_MASK)
	    != InputEvent.BUTTON1_MASK) {
	  m_OutText.selectAll();
	}
      }
    });
    m_History.setBorder(BorderFactory.createTitledBorder("Result list (right-click for options)"));
    m_ClassifierEditor.setClassType(Classifier.class);
    m_ClassifierEditor.setValue(new weka.classifiers.rules.ZeroR());
    m_ClassifierEditor.addPropertyChangeListener(new PropertyChangeListener() {
      public void propertyChange(PropertyChangeEvent e) {
	repaint();
      }
    });

    m_ClassCombo.setToolTipText("Select the attribute to use as the class");
    m_TrainBut.setToolTipText("Test on the same set that the classifier"
			      + " is trained on");
    m_CVBut.setToolTipText("Perform a n-fold cross-validation");
    m_PercentBut.setToolTipText("Train on a percentage of the data and"
				+ " test on the remainder");
    m_TestSplitBut.setToolTipText("Test on a user-specified dataset");
    m_StartBut.setToolTipText("Starts the classification");
    m_StopBut.setToolTipText("Stops a running classification");
    m_StorePredictionsBut.
      setToolTipText("Store predictions in the result list for later "
		     +"visualization");
    m_OutputModelBut
      .setToolTipText("Output the model obtained from the full training set");
    m_OutputPerClassBut.setToolTipText("Output precision/recall & true/false"
				    + " positives for each class");
    m_OutputConfusionBut
      .setToolTipText("Output the matrix displaying class confusions");
    m_OutputEntropyBut
      .setToolTipText("Output entropy-based evaluation measures");
    m_EvalWRTCostsBut
      .setToolTipText("Evaluate errors with respect to a cost matrix");
    m_OutputPredictionsTextBut
      .setToolTipText("Include the predictions on the test set in the output buffer");

    m_FileChooser.setFileFilter(m_ModelFilter);
    m_FileChooser.setFileSelectionMode(JFileChooser.FILES_ONLY);

    m_StorePredictionsBut.setSelected(true);
    m_OutputModelBut.setSelected(true);
    m_OutputPerClassBut.setSelected(true);
    m_OutputConfusionBut.setSelected(true);
    m_ClassCombo.setEnabled(false);
    m_ClassCombo.setPreferredSize(COMBO_SIZE);
    m_ClassCombo.setMaximumSize(COMBO_SIZE);
    m_ClassCombo.setMinimumSize(COMBO_SIZE);

    m_CVBut.setSelected(true);
    updateRadioLinks();
    ButtonGroup bg = new ButtonGroup();
    bg.add(m_TrainBut);
    bg.add(m_CVBut);
    bg.add(m_PercentBut);
    bg.add(m_TestSplitBut);
    m_TrainBut.addActionListener(m_RadioListener);
    m_CVBut.addActionListener(m_RadioListener);
    m_PercentBut.addActionListener(m_RadioListener);
    m_TestSplitBut.addActionListener(m_RadioListener);
    m_SetTestBut.addActionListener(new ActionListener() {
      public void actionPerformed(ActionEvent e) {
	setTestSet();
      }