/* * 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. */ /* * ThresholdVisualizePanel.java * Copyright (C) 2003 University of Waikato, Hamilton, New Zealand * */ package weka.gui.visualize; import weka.classifiers.Classifier; import weka.classifiers.AbstractClassifier; import weka.classifiers.AbstractClassifier; import weka.classifiers.evaluation.EvaluationUtils; import weka.classifiers.evaluation.ThresholdCurve; import weka.core.FastVector; import weka.core.Instances; import weka.core.SingleIndex; import weka.core.Utils; import java.awt.BorderLayout; import java.awt.event.ActionEvent; import java.awt.event.ActionListener; import java.awt.event.WindowAdapter; import java.awt.event.WindowEvent; import java.io.BufferedReader; import java.io.FileReader; import javax.swing.BorderFactory; import javax.swing.JFrame; import javax.swing.border.TitledBorder; /** * This panel is a VisualizePanel, with the added ablility to display the * area under the ROC curve if an ROC curve is chosen. * * @author Dale Fletcher (dale@cs.waikato.ac.nz) * @author FracPete (fracpete at waikato dot ac dot nz) * @version $Revision: 5928 $ */ public class ThresholdVisualizePanel extends VisualizePanel { /** for serialization */ private static final long serialVersionUID = 3070002211779443890L; /** The string to add to the Plot Border. */ private String m_ROCString=""; /** Original border text */ private String m_savePanelBorderText; /** * default constructor */ public ThresholdVisualizePanel() { super(); // Save the current border text TitledBorder tb=(TitledBorder) m_plotSurround.getBorder(); m_savePanelBorderText = tb.getTitle(); } /** * Set the string with ROC area * @param str ROC area string to add to border */ public void setROCString(String str) { m_ROCString=str; } /** * This extracts the ROC area string * @return ROC area string */ public String getROCString() { return m_ROCString; } /** * This overloads VisualizePanel's setUpComboBoxes to add * ActionListeners to watch for when the X/Y Axis comboboxes * are changed. * @param inst a set of instances with data for plotting */ public void setUpComboBoxes(Instances inst) { super.setUpComboBoxes(inst); m_XCombo.addActionListener(new ActionListener() { public void actionPerformed(ActionEvent e) { setBorderText(); } }); m_YCombo.addActionListener(new ActionListener() { public void actionPerformed(ActionEvent e) { setBorderText(); } }); // Just in case the default is ROC setBorderText(); } /** * This checks the current selected X/Y Axis comboBoxes to see if * an ROC graph is selected. If so, add the ROC area string to the * plot border, otherwise display the original border text. */ private void setBorderText() { String xs = m_XCombo.getSelectedItem().toString(); String ys = m_YCombo.getSelectedItem().toString(); if (xs.equals("X: False Positive Rate (Num)") && ys.equals("Y: True Positive Rate (Num)")) { m_plotSurround.setBorder((BorderFactory.createTitledBorder(m_savePanelBorderText+" "+m_ROCString))); } else m_plotSurround.setBorder((BorderFactory.createTitledBorder(m_savePanelBorderText))); } /** * displays the previously saved instances * * @param insts the instances to display * @throws Exception if display is not possible */ protected void openVisibleInstances(Instances insts) throws Exception { super.openVisibleInstances(insts); setROCString( "(Area under ROC = " + Utils.doubleToString(ThresholdCurve.getROCArea(insts), 4) + ")"); setBorderText(); } /** * Starts the ThresholdVisualizationPanel with parameters from the command line.

* * Valid options are:

* -h
* lists all the commandline parameters

* * -t file
* Dataset to process with given classifier.

* * -W classname
* Full classname of classifier to run.
* Options after '--' are passed to the classifier.
* (default weka.classifiers.functions.Logistic)

* * -r number
* The number of runs to perform (default 2).

* * -x number
* The number of Cross-validation folds (default 10).

* * -l file
* Previously saved threshold curve ARFF file.

* * @param args optional commandline parameters */ public static void main(String [] args) { Instances inst; Classifier classifier; int runs; int folds; String tmpStr; boolean compute; Instances result; String[] options; SingleIndex classIndex; SingleIndex valueIndex; int seed; inst = null; classifier = null; runs = 2; folds = 10; compute = true; result = null; classIndex = null; valueIndex = null; seed = 1; try { // help? if (Utils.getFlag('h', args)) { System.out.println("\nOptions for " + ThresholdVisualizePanel.class.getName() + ":\n"); System.out.println("-h\n\tThis help."); System.out.println("-t \n\tDataset to process with given classifier."); System.out.println("-c \n\tThe class index. first and last are valid, too (default: last)."); System.out.println("-C \n\tThe index of the class value to get the the curve for (default: first)."); System.out.println("-W \n\tFull classname of classifier to run.\n\tOptions after '--' are passed to the classifier.\n\t(default: weka.classifiers.functions.Logistic)"); System.out.println("-r \n\tThe number of runs to perform (default: 1)."); System.out.println("-x \n\tThe number of Cross-validation folds (default: 10)."); System.out.println("-S \n\tThe seed value for randomizing the data (default: 1)."); System.out.println("-l \n\tPreviously saved threshold curve ARFF file."); return; } // regular options tmpStr = Utils.getOption('l', args); if (tmpStr.length() != 0) { result = new Instances(new BufferedReader(new FileReader(tmpStr))); compute = false; } if (compute) { tmpStr = Utils.getOption('r', args); if (tmpStr.length() != 0) runs = Integer.parseInt(tmpStr); else runs = 1; tmpStr = Utils.getOption('x', args); if (tmpStr.length() != 0) folds = Integer.parseInt(tmpStr); else folds = 10; tmpStr = Utils.getOption('S', args); if (tmpStr.length() != 0) seed = Integer.parseInt(tmpStr); else seed = 1; tmpStr = Utils.getOption('t', args); if (tmpStr.length() != 0) { inst = new Instances(new BufferedReader(new FileReader(tmpStr))); inst.setClassIndex(inst.numAttributes() - 1); } tmpStr = Utils.getOption('W', args); if (tmpStr.length() != 0) { options = Utils.partitionOptions(args); } else { tmpStr = weka.classifiers.functions.Logistic.class.getName(); options = new String[0]; } classifier = AbstractClassifier.forName(tmpStr, options); tmpStr = Utils.getOption('c', args); if (tmpStr.length() != 0) classIndex = new SingleIndex(tmpStr); else classIndex = new SingleIndex("last"); tmpStr = Utils.getOption('C', args); if (tmpStr.length() != 0) valueIndex = new SingleIndex(tmpStr); else valueIndex = new SingleIndex("first"); } // compute if necessary if (compute) { if (classIndex != null) { classIndex.setUpper(inst.numAttributes() - 1); inst.setClassIndex(classIndex.getIndex()); } else { inst.setClassIndex(inst.numAttributes() - 1); } if (valueIndex != null) { valueIndex.setUpper(inst.classAttribute().numValues() - 1); } ThresholdCurve tc = new ThresholdCurve(); EvaluationUtils eu = new EvaluationUtils(); FastVector predictions = new FastVector(); for (int i = 0; i < runs; i++) { eu.setSeed(seed + i); predictions.appendElements(eu.getCVPredictions(classifier, inst, folds)); } if (valueIndex != null) result = tc.getCurve(predictions, valueIndex.getIndex()); else result = tc.getCurve(predictions); } // setup GUI ThresholdVisualizePanel vmc = new ThresholdVisualizePanel(); vmc.setROCString("(Area under ROC = " + Utils.doubleToString(ThresholdCurve.getROCArea(result), 4) + ")"); if (compute) vmc.setName( result.relationName() + ". (Class value " + inst.classAttribute().value(valueIndex.getIndex()) + ")"); else vmc.setName( result.relationName() + " (display only)"); PlotData2D tempd = new PlotData2D(result); tempd.setPlotName(result.relationName()); tempd.addInstanceNumberAttribute(); vmc.addPlot(tempd); String plotName = vmc.getName(); final JFrame jf = new JFrame("Weka Classifier Visualize: "+plotName); jf.setSize(500,400); jf.getContentPane().setLayout(new BorderLayout()); jf.getContentPane().add(vmc, BorderLayout.CENTER); jf.addWindowListener(new WindowAdapter() { public void windowClosing(WindowEvent e) { jf.dispose(); } }); jf.setVisible(true); } catch (Exception e) { e.printStackTrace(); } } }