classifierpanel.java

:<<数据挖掘--实用机器学习技术及java实现>>一书的配套源程序

/*
 *    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 java.io.*;
import java.util.LinkedList;
import java.net.*;

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.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 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.ScrollPaneConstants;
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.OverlayLayout;


/**
 * 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 Dave Musicant ([e-mail])
 * @author Sebastian Celis ([e-mail])
 * @version $Revision: 1.45 $
 */
public class ClassifierPanel extends JPanel {

  /**
   * This is used to describe the status of each fold when they are sent
   * out to other computers.
   */
  interface Status
  {
      int
          NOT_DONE = 0,
          DONE = 1;
  }

  /**
   * When connecting to one of the other computers to give it data, this
   * program must first identify if it is connecting through Weka's CLI or
   * GUI options.
   */
  interface Connection
  {
      int
          CV_CLI = 0,
          CV_GUI = 1;
  }

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

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

  /** Click to set the settings for cross-validation */
  protected JButton m_CVSettingsBut = new JButton("Settings...");

  /** Check to have the cross-validation step run in parallel */
  protected JCheckBox m_RunInParallelBut =
    new JCheckBox("Run in parallel");

  /** Click to edit the settings of running cross-validation in parallel */
  protected JButton m_ParallelConfigBut =
    new JButton("Configure parallelization...");

  /** Label by where the port number is entered */
  protected JLabel m_PortLab = new JLabel("Port Number", SwingConstants.RIGHT);

  /** TextField where the port number is entered */
  protected JTextField m_PortText = new JTextField();

  /**
   * Area where all the computers to connect to during parallel
   * cross validation are entered
   */
  protected JTextArea m_AddressesArea
      = new JTextArea(10, 25);

  /** 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...");

  /** 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 preprocess panel through which filters can be applied to
      user supplied test data sets */
  protected PreprocessPanel m_Preprocess = null;

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

  /**
   * This is the client that will connect to all of the other computers,
   * if parallel cross-validation is taking place, hand off data, and
   * expect results back.
   *
   * @author Dave Musicant ([e-mail])
   * @author Sebastian Celis ([e-mail])
   */
  private class GUIDistributedClient
  {
      /** The port the client will connect on */
      private int port;

      /**
       * A list of strings containing the names of the computers to connect
       * to
       */
      private LinkedList computers;

      /** The predicted instances */
      private Instances predInstances;

      /** The actual data cross-validation is to be done on */
      private Instances inst;

      /** Tells the status of each fold and whether it is DONE or NOT_DONE */
      private int status[];

      /** The number of folds to run */
      private int numFolds;

      /** The number of folds completed */
      private int foldsCompleted;

      /** The index of the last fold calculated. */
      private int lastIndexSent;

      /** The classifier used in the cross-validation. */
      private Classifier classifier;