Examples of weka.classifiers.evaluation.ThresholdCurve

weka.classifiers.evaluation.ThresholdCurve
Generates points illustrating prediction tradeoffs that can be obtained by varying the threshold value between classes. For example, the typical threshold value of 0.5 means the predicted probability of "positive" must be higher than 0.5 for the instance to be predicted as "positive". The resulting dataset can be used to visualize precision/recall tradeoff, or for ROC curve analysis (true positive rate vs false positive rate). Weka just varies the threshold on the class probability estimates in each case. The Mann Whitney statistic is used to calculate the AUC. @author Len Trigg (len@reeltwo.com) @version $Revision: 5987 $


    // Check if any predictions have been collected
    if (m_Predictions == null) {
      return Utils.missingValue();
    } else {
      ThresholdCurve tc = new ThresholdCurve();
      Instances result = tc.getCurve(m_Predictions, classIndex);
      return ThresholdCurve.getROCArea(result);
    }
  }

View Full Code Here

      if (thresholdLabel.length() != 0)
        labelIndex = template.classAttribute().indexOfValue(thresholdLabel);
      if (labelIndex == -1)
        throw new IllegalArgumentException(
            "Class label '" + thresholdLabel + "' is unknown!");
      ThresholdCurve tc = new ThresholdCurve();
      Instances result = tc.getCurve(testingEvaluation.predictions(), labelIndex);
      DataSink.write(thresholdFile, result);
    }


    return text.toString();
  }

View Full Code Here

                  }
                  


      if (ce.getTestSet().getDataSet().classAttribute().isNominal() &&
          m_thresholdListeners.size() > 0) {
        ThresholdCurve tc = new ThresholdCurve();
        Instances result = tc.getCurve(m_eval.predictions(), 0);
        result.
          setRelationName(ce.getTestSet().getDataSet().relationName());
        PlotData2D pd = new PlotData2D(result);
        String htmlTitle = "<html><font size=-2>"
          + textTitle;

View Full Code Here

  JMenuItem clv = new JMenuItem(classAtt.value(i));
  final int classValue = i;
  clv.addActionListener(new ActionListener() {
      public void actionPerformed(ActionEvent e) {
        try {
    ThresholdCurve tc = new ThresholdCurve();
    Instances result = tc.getCurve(preds, classValue);
    //VisualizePanel vmc = new VisualizePanel();
    ThresholdVisualizePanel vmc = new ThresholdVisualizePanel();
    vmc.setROCString("(Area under ROC = " + 
         Utils.doubleToString(ThresholdCurve.getROCArea(result), 4) + ")");
    vmc.setLog(m_Log);
    vmc.setName(result.relationName()+". (Class value "+
          classAtt.value(classValue)+")");
    PlotData2D tempd = new PlotData2D(result);
    tempd.setPlotName(result.relationName());
    tempd.addInstanceNumberAttribute();
    // specify which points are connected
    boolean[] cp = new boolean[result.numInstances()];
    for (int n = 1; n < cp.length; n++)
      cp[n] = true;
    tempd.setConnectPoints(cp);
    // add plot
    vmc.addPlot(tempd);
    visualizeClassifierErrors(vmc);
        } catch (Exception ex) {
    ex.printStackTrace();
        }
        }
    });
    visThreshold.add(clv);
      }
    } else {
      visThreshold.setEnabled(false);
    }
    resultListMenu.add(visThreshold);
    
    JMenu visCostBenefit = new JMenu("Cost/Benefit analysis");
    if ((preds != null) && (classAtt != null) && (classAtt.isNominal())) {
      for (int i = 0; i < classAtt.numValues(); i++) {
        JMenuItem clv = new JMenuItem(classAtt.value(i));
        final int classValue = i;
        clv.addActionListener(new ActionListener() {
            public void actionPerformed(ActionEvent e) {
              try {
                ThresholdCurve tc = new ThresholdCurve();
                Instances result = tc.getCurve(preds, classValue);


                // Create a dummy class attribute with the chosen
                // class value as index 0 (if necessary).
                Attribute classAttToUse = classAtt;
                if (classValue != 0) {

View Full Code Here

        }




        if (testData.classAttribute().isNominal() &&
            m_thresholdListeners.size() > 0) {
          ThresholdCurve tc = new ThresholdCurve();
          Instances result = tc.getCurve(m_eval.predictions(), 0);
          result.
          setRelationName(testData.relationName());
          PlotData2D pd = new PlotData2D(result);
          String htmlTitle = "<html><font size=-2>"
            + textTitle;

View Full Code Here


    // Check if any predictions have been collected
    if (m_Predictions == null) {
      return Instance.missingValue();
    } else {
      ThresholdCurve tc = new ThresholdCurve();
      Instances result = tc.getCurve(m_Predictions, classIndex);
      return ThresholdCurve.getROCArea(result);
    }
  }

View Full Code Here

      if (thresholdLabel.length() != 0)
  labelIndex = template.classAttribute().indexOfValue(thresholdLabel);
      if (labelIndex == -1)
  throw new IllegalArgumentException(
      "Class label '" + thresholdLabel + "' is unknown!");
      ThresholdCurve tc = new ThresholdCurve();
      Instances result = tc.getCurve(testingEvaluation.predictions(), labelIndex);
      DataSink.write(thresholdFile, result);
    }
    
    return text.toString();
  }

View Full Code Here

                  }
                  


      if (ce.getTestSet().getDataSet().classAttribute().isNominal() &&
          m_thresholdListeners.size() > 0) {
        ThresholdCurve tc = new ThresholdCurve();
        Instances result = tc.getCurve(m_eval.predictions(), 0);
        result.
          setRelationName(ce.getTestSet().getDataSet().relationName());
        PlotData2D pd = new PlotData2D(result);
        String htmlTitle = Messages.getInstance().getString("ClassifierPerformanceEvaluator_AcceptClassifier_HtmlTitle_Text_First")
          + textTitle;

View Full Code Here

  JMenuItem clv = new JMenuItem(classAtt.value(i));
  final int classValue = i;
  clv.addActionListener(new ActionListener() {
      public void actionPerformed(ActionEvent e) {
        try {
    ThresholdCurve tc = new ThresholdCurve();
    Instances result = tc.getCurve(preds, classValue);
    //VisualizePanel vmc = new VisualizePanel();
    ThresholdVisualizePanel vmc = new ThresholdVisualizePanel();
    vmc.setROCString(Messages.getInstance().getString("ClassifierPanel_Visualize_VMC_SetROCString_Text_First") + 
         Utils.doubleToString(ThresholdCurve.getROCArea(result), 4) + Messages.getInstance().getString("ClassifierPanel_Visualize_VMC_SetROCString_Text_Second"));
    vmc.setLog(m_Log);
    vmc.setName(result.relationName()+Messages.getInstance().getString("ClassifierPanel_Visualize_VMC_SetName_Text_First") +
          classAtt.value(classValue) + Messages.getInstance().getString("ClassifierPanel_Visualize_VMC_SetName_Text_Second"));
    PlotData2D tempd = new PlotData2D(result);
    tempd.setPlotName(result.relationName());
    tempd.addInstanceNumberAttribute();
    // specify which points are connected
    boolean[] cp = new boolean[result.numInstances()];
    for (int n = 1; n < cp.length; n++)
      cp[n] = true;
    tempd.setConnectPoints(cp);
    // add plot
    vmc.addPlot(tempd);
    visualizeClassifierErrors(vmc);
        } catch (Exception ex) {
    ex.printStackTrace();
        }
        }
    });
    visThreshold.add(clv);
      }
    } else {
      visThreshold.setEnabled(false);
    }
    resultListMenu.add(visThreshold);
    
    JMenu visCostBenefit = new JMenu(Messages.getInstance().getString("ClassifierPanel_Visualize_VisCostBenefit_JMenu_Text"));
    if ((preds != null) && (classAtt != null) && (classAtt.isNominal())) {
      for (int i = 0; i < classAtt.numValues(); i++) {
        JMenuItem clv = new JMenuItem(classAtt.value(i));
        final int classValue = i;
        clv.addActionListener(new ActionListener() {
            public void actionPerformed(ActionEvent e) {
              try {
                ThresholdCurve tc = new ThresholdCurve();
                Instances result = tc.getCurve(preds, classValue);


                // Create a dummy class attribute with the chosen
                // class value as index 0 (if necessary).
                Attribute classAttToUse = classAtt;
                if (classValue != 0) {

View Full Code Here

  
  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(Messages.getInstance().getString("ThresholdVisualizePanel_OpenVisibleInstances_Text_Third") +

View Full Code Here

0 1 2

TOP

Related Classes of weka.classifiers.evaluation.ThresholdCurve

bnGUI.gui_WekaExperiment

bnGUI.Thread_weka_run

weka.classifiers.Classifier

weka.classifiers.Evaluation

weka.classifiers.meta.ThresholdSelector

weka.core.Attribute

weka.core.DenseInstance

weka.core.FastVector

weka.core.Instance

weka.core.Instances

All source code are property of their respective owners. Java is a trademark of Sun Microsystems, Inc and owned by ORACLE Inc. Contact coftware#gmail.com.