clustererpanel.java

MacroWeka扩展了著名数据挖掘工具weka

package chen.macroweka.gui.explorer;

import weka.core.Instances;
import weka.core.Instance;
import weka.core.OptionHandler;
import weka.core.Attribute;
import weka.core.Utils;
import weka.core.FastVector;
import weka.core.SerializedObject;
import weka.core.Drawable;
import weka.clusterers.Clusterer;
import weka.clusterers.ClusterEvaluation;
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.InstancesSummaryPanel;
import weka.gui.SaveBuffer;
import weka.filters.Filter;
import weka.filters.unsupervised.attribute.Remove;
import weka.gui.visualize.VisualizePanel;
import weka.gui.visualize.PlotData2D;
import weka.gui.visualize.Plot2D;
import weka.gui.treevisualizer.*;
import weka.gui.ListSelectorDialog;
import weka.gui.ExtensionFileFilter;

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.Dimension;
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.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 java.awt.Point;
import javax.swing.JPopupMenu;
import javax.swing.JMenu;
import javax.swing.JMenuItem;
import javax.swing.DefaultListModel;
import javax.swing.JList;
import javax.swing.filechooser.FileFilter;

/** 
 * This panel allows the user to select and configure a clusterer, and evaluate
 * the clusterer using a number of testing modes (test on the training data,
 * train/test on a percentage split, test on a
 * separate split). The results of clustering runs are stored in a result
 * history so that previous results are accessible.
 *
 * @author Mark Hall ([e-mail])
 * @author Richard Kirkby ([e-mail])
 * @version $Revision: 1.1 $
 */
public class ClustererPanel extends JPanel {

  /** The filename extension that should be used for model files */
  public static String MODEL_FILE_EXTENSION = ".model";

  /** Lets the user configure the clusterer */
  protected GenericObjectEditor m_ClustererEditor =
    new GenericObjectEditor();

  /** The panel showing the current clusterer selection */
  protected PropertyPanel m_CLPanel = new PropertyPanel(m_ClustererEditor);
  
  /** 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);

  /** 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");

  /** Click to set test mode to classes to clusters based evaluation */
  protected JRadioButton m_ClassesToClustersBut = 
    new JRadioButton("Classes to clusters evaluation");

  /** Lets the user select the class column for classes to clusters based
      evaluation */
  protected JComboBox m_ClassCombo = new JComboBox();

  /** 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 button used to popup a list for choosing attributes to ignore while
      clustering */
  protected JButton m_ignoreBut = new JButton("Ignore attributes");

  protected DefaultListModel m_ignoreKeyModel = new DefaultListModel();
  protected JList m_ignoreKeyList = new JList(m_ignoreKeyModel);

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

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

  /** Stop the class combo from taking up to much space */
  private Dimension COMBO_SIZE = new Dimension(250, m_StartBut
					       .getPreferredSize().height);

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

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

  /** Check to save the predictions in the results list for visualizing
      later on */
  protected JCheckBox m_StorePredictionsBut = 
    new JCheckBox("Store clusters for visualization");
  
  /** A thread that clustering runs in */
  protected Thread m_RunThread;
  
  /** The instances summary panel displayed by m_SetTestFrame */
  protected InstancesSummaryPanel m_Summary;

  /** Filter to ensure only model files are selected */  
  protected FileFilter m_ModelFilter =
    new ExtensionFileFilter(MODEL_FILE_EXTENSION, "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 {
    GenericObjectEditor.registerEditors();
  }
  
  /**
   * Creates the clusterer panel
   */
  public ClustererPanel() {

    // 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_ClustererEditor.setClassType(Clusterer.class);
    m_ClustererEditor.setValue(new weka.clusterers.EM());
    m_ClustererEditor.addPropertyChangeListener(new PropertyChangeListener() {
      public void propertyChange(PropertyChangeEvent e) {
	repaint();
      }
    });

    m_TrainBut.setToolTipText("Cluster the same set that the clusterer"
			      + " is trained on");
    m_PercentBut.setToolTipText("Train on a percentage of the data and"
				+ " cluster the remainder");
    m_TestSplitBut.setToolTipText("Cluster a user-specified dataset");
    m_ClassesToClustersBut.setToolTipText("Evaluate clusters with respect to a"
					  +" class");
    m_ClassCombo.setToolTipText("Select the class attribute for class based"
				+" evaluation");
    m_StartBut.setToolTipText("Starts the clustering");
    m_StopBut.setToolTipText("Stops a running clusterer");
    m_StorePredictionsBut.
      setToolTipText("Store predictions in the result list for later "
		     +"visualization");
    m_ignoreBut.setToolTipText("Ignore attributes during clustering");

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

    m_ClassCombo.setPreferredSize(COMBO_SIZE);
    m_ClassCombo.setMaximumSize(COMBO_SIZE);
    m_ClassCombo.setMinimumSize(COMBO_SIZE);
    m_ClassCombo.setEnabled(false);

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

    m_StartBut.setEnabled(false);
    m_StopBut.setEnabled(false);
    m_ignoreBut.setEnabled(false);
    m_StartBut.addActionListener(new ActionListener() {
      public void actionPerformed(ActionEvent e) {
	startClusterer();
      }
    });
    m_StopBut.addActionListener(new ActionListener() {
      public void actionPerformed(ActionEvent e) {
	stopClusterer();
      }
    });

    m_ignoreBut.addActionListener(new ActionListener() {
	public void actionPerformed(ActionEvent e) {
	  setIgnoreColumns();
	}
      });
   
    m_History.setHandleRightClicks(false);
    // see if we can popup a menu for the selected result
    m_History.getList().addMouseListener(new MouseAdapter() {
	public void mouseClicked(MouseEvent e) {
	  if (((e.getModifiers() & InputEvent.BUTTON1_MASK)
	       != InputEvent.BUTTON1_MASK) || e.isAltDown()) {
	    int index = m_History.getList().locationToIndex(e.getPoint());
	    if (index != -1) {
	      String name = m_History.getNameAtIndex(index);
	      visualizeClusterer(name, e.getX(), e.getY());
	    } else {
	      visualizeClusterer(null, e.getX(), e.getY());
	    }
	  }
	}
      });

    // Layout the GUI
    JPanel p1 = new JPanel();
    p1.setBorder(BorderFactory.createCompoundBorder(
		 BorderFactory.createTitledBorder("Clusterer"),
		 BorderFactory.createEmptyBorder(0, 5, 5, 5)
		 ));
    p1.setLayout(new BorderLayout());
    p1.add(m_CLPanel, BorderLayout.NORTH);

    JPanel p2 = new JPanel();
    GridBagLayout gbL = new GridBagLayout();
    p2.setLayout(gbL);
    p2.setBorder(BorderFactory.createCompoundBorder(
		 BorderFactory.createTitledBorder("Cluster mode"),
		 BorderFactory.createEmptyBorder(0, 5, 5, 5)
		 ));
    GridBagConstraints gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.WEST;
    gbC.gridy = 0;     gbC.gridx = 0;
    gbL.setConstraints(m_TrainBut, gbC);
    p2.add(m_TrainBut);

    gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.WEST;
    gbC.gridy = 1;     gbC.gridx = 0;
    gbL.setConstraints(m_TestSplitBut, gbC);
    p2.add(m_TestSplitBut);

    gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.EAST;
    gbC.fill = GridBagConstraints.HORIZONTAL;
    gbC.gridy = 1;     gbC.gridx = 1;    gbC.gridwidth = 2;
    gbC.insets = new Insets(2, 10, 2, 0);
    gbL.setConstraints(m_SetTestBut, gbC);
    p2.add(m_SetTestBut);

    gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.WEST;
    gbC.gridy = 2;     gbC.gridx = 0;
    gbL.setConstraints(m_PercentBut, gbC);
    p2.add(m_PercentBut);

    gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.EAST;
    gbC.fill = GridBagConstraints.HORIZONTAL;
    gbC.gridy = 2;     gbC.gridx = 1;
    gbC.insets = new Insets(2, 10, 2, 10);
    gbL.setConstraints(m_PercentLab, gbC);
    p2.add(m_PercentLab);

    gbC = new GridBagConstraints();
    gbC.anchor = GridBagConstraints.EAST;