Examples of weka.clusterers.ClusterEvaluation

weka.clusterers.ClusterEvaluation
Class for evaluating clustering models.
Valid options are:
-t name of the training file
Specify the training file.
-T name of the test file
Specify the test file to apply clusterer to.
-d name of file to save clustering model to
Specify output file.
-l name of file to load clustering model from
Specifiy input file.
-p attribute range
Output predictions. Predictions are for the training file if only the training file is specified, otherwise they are for the test file. The range specifies attribute values to be output with the predictions. Use '-p 0' for none.
-x num folds
Set the number of folds for a cross validation of the training data. Cross validation can only be done for distribution clusterers and will be performed if the test file is missing.
-s num
Sets the seed for randomizing the data for cross-validation.
-c class
Set the class attribute. If set, then class based evaluation of clustering is performed.
-g name of graph file
Outputs the graph representation of the clusterer to the file. Only for clusterer that implemented the weka.core.Drawable interface.
@author Mark Hall (mhall@cs.waikato.ac.nz) @version $Revision: 7753 $ @see weka.core.Drawable

      }
      // copy full model for output
      SerializedObject so = new SerializedObject(clusterer);
      fullClusterer = (Clusterer) so.getObject();
      
      ClusterEvaluation eval = new ClusterEvaluation();
      eval.setClusterer(clusterer);
      switch (testMode) {
        case 3: case 5: // Test on training
        m_Log.statusMessage("Clustering training data...");
        eval.evaluateClusterer(trainInst);
        plotInstances.setInstances(inst);
        plotInstances.setClusterEvaluation(eval);
        outBuff.append("=== Model and evaluation on training set ===\n\n");
        break;


        case 2: // Percent split
        m_Log.statusMessage("Randomizing instances...");
        inst.randomize(new Random(1));
        trainInst.randomize(new Random(1));
        int trainSize = trainInst.numInstances() * percent / 100;
        int testSize = trainInst.numInstances() - trainSize;
        Instances train = new Instances(trainInst, 0, trainSize);
        Instances test = new Instances(trainInst, trainSize, testSize);
        Instances testVis = new Instances(inst, trainSize, testSize);
        m_Log.statusMessage("Building model on training split...");
        clusterer.buildClusterer(train);
        m_Log.statusMessage("Evaluating on test split...");
        eval.evaluateClusterer(test);
        plotInstances.setInstances(testVis);
        plotInstances.setClusterEvaluation(eval);
        outBuff.append("=== Model and evaluation on test split ===\n");
        break;
    
        case 4: // Test on user split
        m_Log.statusMessage("Evaluating on test data...");
        Instances userTestT = new Instances(userTest);
        if (!m_ignoreKeyList.isSelectionEmpty()) {
    userTestT = removeIgnoreCols(userTestT);
        }
        eval.evaluateClusterer(userTestT);
        plotInstances.setInstances(userTest);
        plotInstances.setClusterEvaluation(eval);
        outBuff.append("=== Model and evaluation on test set ===\n");
        break;


        default:
        throw new Exception("Test mode not implemented");
      }
      outBuff.append(eval.clusterResultsToString());
      outBuff.append("\n");
      m_History.updateResult(name);
      m_Log.logMessage("Finished " + cname);
      m_Log.statusMessage("OK");
    } catch (Exception ex) {

View Full Code Here


              m_Log.logMessage("Started reevaluate model");
              if (m_Log instanceof TaskLogger) {
                ((TaskLogger)m_Log).taskStarted();
              }
              ClusterEvaluation eval = new ClusterEvaluation();
              eval.setClusterer(clusterer);
    
              Instances userTestT = new Instances(userTest);
              if (ignoredAtts != null) {
                userTestT = removeIgnoreCols(userTestT, ignoredAtts);
              }


              eval.evaluateClusterer(userTestT);
      
              plotInstances.setClusterEvaluation(eval);
              plotInstances.setInstances(userTest);
              plotInstances.setUp();


              outBuff.append("\n=== Re-evaluation on test set ===\n\n");
              outBuff.append("User supplied test set\n");  
              outBuff.append("Relation:     " + userTest.relationName() + '\n');
              outBuff.append("Instances:    " + userTest.numInstances() + '\n');
              outBuff.append("Attributes:   " + userTest.numAttributes() + "\n\n");
              if (trainHeader == null)
                outBuff.append("NOTE - if test set is not compatible then results are "
                               + "unpredictable\n\n");
      
              outBuff.append(eval.clusterResultsToString());
              outBuff.append("\n");
              m_History.updateResult(name);
              m_Log.logMessage("Finished re-evaluation");
              m_Log.statusMessage("OK");
            } catch (Exception ex) {

View Full Code Here

    }
    train.setClassIndex(-1);
    test.setClassIndex(-1);
      


    ClusterEvaluation eval = new ClusterEvaluation();


    Object [] result = new Object[overall_length];
    long trainTimeStart = System.currentTimeMillis();
    m_clusterer.buildClusterer(train);
    double numClusters = m_clusterer.numberOfClusters();
    eval.setClusterer(m_clusterer);
    long trainTimeElapsed = System.currentTimeMillis() - trainTimeStart;
    long testTimeStart = System.currentTimeMillis();
    eval.evaluateClusterer(test);
    long testTimeElapsed = System.currentTimeMillis() - testTimeStart;
    //    m_result = eval.toSummaryString();


    // The results stored are all per instance -- can be multiplied by the
    // number of instances to get absolute numbers
    int current = 0;
    result[current++] = new Double(train.numInstances());
    result[current++] = new Double(test.numInstances());


    result[current++] = new Double(eval.getLogLikelihood());
    result[current++] = new Double(numClusters);
    
    // Timing stats
    result[current++] = new Double(trainTimeElapsed / 1000.0);
    result[current++] = new Double(testTimeElapsed / 1000.0);

View Full Code Here

      }
      // copy full model for output
      SerializedObject so = new SerializedObject(clusterer);
      fullClusterer = (Clusterer) so.getObject();
      
      ClusterEvaluation eval = new ClusterEvaluation();
      eval.setClusterer(clusterer);
      switch (testMode) {
        case 3: case 5: // Test on training
        m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Third"));
        eval.evaluateClusterer(trainInst, "", false);
        predData = setUpVisualizableInstances(inst,eval);
        outBuff.append(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_OutBuffer_Text_TwentySecond"));
        break;


        case 2: // Percent split
        m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Fourth"));
        inst.randomize(new Random(1));
        trainInst.randomize(new Random(1));
        int trainSize = trainInst.numInstances() * percent / 100;
        int testSize = trainInst.numInstances() - trainSize;
        Instances train = new Instances(trainInst, 0, trainSize);
        Instances test = new Instances(trainInst, trainSize, testSize);
        Instances testVis = new Instances(inst, trainSize, testSize);
        m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Fifth"));
        trainTimeStart = System.currentTimeMillis();
        clusterer.buildClusterer(train);
        trainTimeElapsed = System.currentTimeMillis() - trainTimeStart;
        
        m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Sixth"));
        eval.evaluateClusterer(test, "", false);
        predData = setUpVisualizableInstances(testVis, eval);
        outBuff.append(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_OutBuffer_Text_TwentyThird"));
        outBuff.append(clusterer.toString() + '\n');
              outBuff.append(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_OutBuffer_Text_TimeTakenPercentage") +
                  Utils.doubleToString(trainTimeElapsed / 1000.0,2)
                  + " " + Messages.getInstance().getString("ClassifierPanel_StartClassifier_OutBuffer_Text_TwentyNineth"));
        break;
    
        case 4: // Test on user split
        m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Seventh"));
        Instances userTestT = new Instances(userTest);
        if (!m_ignoreKeyList.isSelectionEmpty()) {
    userTestT = removeIgnoreCols(userTestT);
        }
        eval.evaluateClusterer(userTestT, "", false);
        predData = setUpVisualizableInstances(userTest, eval);
        outBuff.append(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_OutBuffer_Text_TwentyFourth"));
        break;


        default:
        throw new Exception(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Exception_Text_Sixth"));
      }
      outBuff.append(eval.clusterResultsToString());
      outBuff.append("\n");
      m_History.updateResult(name);
      m_Log.logMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_LogMessage_Text_Third") + cname);
      m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_StartClusterer_Run_Log_StatusMessage_Eighth"));
    } catch (Exception ex) {

View Full Code Here


              m_Log.logMessage(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_Log_LogMessage_Text_Third"));
              if (m_Log instanceof TaskLogger) {
                ((TaskLogger)m_Log).taskStarted();
              }
              ClusterEvaluation eval = new ClusterEvaluation();
              eval.setClusterer(clusterer);
    
              Instances userTestT = new Instances(userTest);
              if (ignoredAtts != null) {
                userTestT = removeIgnoreCols(userTestT, ignoredAtts);
              }


              eval.evaluateClusterer(userTestT);
      
              predData = setUpVisualizableInstances(userTest, eval);


              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_First"));
              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Second"));  
              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Third") + userTest.relationName() + Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Fourth"));
              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Fifth") + userTest.numInstances() + Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Sixth"));
              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Seventh") + userTest.numAttributes() + Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Eighth"));
              if (trainHeader == null)
                outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Nineth"));
      
              outBuff.append(eval.clusterResultsToString());
              outBuff.append(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_OutBuffer_Text_Tenth"));
              m_History.updateResult(name);
              m_Log.logMessage(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_Log_LogMessage_Text_Fourth"));
              m_Log.statusMessage(Messages.getInstance().getString("ClustererPanel_ReEvaluateModel_Run_Log_StatusMessage_Text_Third"));
            } catch (Exception ex) {

View Full Code Here

              boolean numericClass = false;  
//        final String oldText = m_visual.getText();
        try {
    if (ce.getSetNumber() == 1 /*|| 
        ce.getClusterer() != m_clusterer */) {
      m_eval = new ClusterEvaluation();
      m_clusterer = ce.getClusterer();
                  m_eval.setClusterer(m_clusterer);
    }
    
    if (ce.getSetNumber() <= ce.getMaxSetNumber()) {

View Full Code Here

   * @param data        the training instances
   * @throws Exception  if something goes wrong
   */
  public void buildClassifier(Instances data) throws Exception {
    Instances    clusterData;
    ClusterEvaluation  eval;
    int      i;
    Instance    instance;
    int[][]     counts;
    int[]     clusterTotals;
    double[]     best;
    double[]     current;
    double[]     clusterAssignments;
    
    // can classifier handle the data?
    getCapabilities().testWithFail(data);


    // remove instances with missing class
    data = new Instances(data);
    data.deleteWithMissingClass();


    // save original header (needed for clusters to classes output)
    m_OriginalHeader = new Instances(data, 0);
    
    // remove class attribute for clusterer
    clusterData = new Instances(data);
    clusterData.setClassIndex(-1);
    clusterData.deleteAttributeAt(m_OriginalHeader.classIndex());
    m_ClusteringHeader = new Instances(clusterData, 0);


    if (m_ClusteringHeader.numAttributes() == 0) {
      System.err.println("Data contains only class attribute, defaulting to ZeroR model.");
      m_ZeroR = new ZeroR();
      m_ZeroR.buildClassifier(data);
    }
    else {
      m_ZeroR = null;
      
      // build clusterer
      m_ActualClusterer = AbstractClusterer.makeCopy(m_Clusterer);
      m_ActualClusterer.buildClusterer(clusterData);


      // evaluate clusterer on training set
      eval = new ClusterEvaluation();
      eval.setClusterer(m_ActualClusterer);
      eval.evaluateClusterer(clusterData);
      clusterAssignments = eval.getClusterAssignments();


      // determine classes-to-clusters mapping
      counts        = new int [eval.getNumClusters()][m_OriginalHeader.numClasses()];
      clusterTotals = new int[eval.getNumClusters()];
      best          = new double[eval.getNumClusters()+1];
      current       = new double[eval.getNumClusters()+1];
      for (i = 0; i < data.numInstances(); i++) {
  instance = data.instance(i);
  counts[(int) clusterAssignments[i]][(int) instance.classValue()]++;
  clusterTotals[(int) clusterAssignments[i]]++;
  i++;
      }
      best[eval.getNumClusters()] = Double.MAX_VALUE;
      ClusterEvaluation.mapClasses(eval.getNumClusters(), 0, counts, clusterTotals, current, best, 0);
      m_ClustersToClasses = new double[best.length];
      System.arraycopy(best, 0, m_ClustersToClasses, 0, best.length);
    }
  }

View Full Code Here

    }
    train.setClassIndex(-1);
    test.setClassIndex(-1);
      


    ClusterEvaluation eval = new ClusterEvaluation();


    Object [] result = new Object[overall_length];
    long trainTimeStart = System.currentTimeMillis();
    m_clusterer.buildClusterer(train);
    double numClusters = m_clusterer.numberOfClusters();
    eval.setClusterer(m_clusterer);
    long trainTimeElapsed = System.currentTimeMillis() - trainTimeStart;
    long testTimeStart = System.currentTimeMillis();
    eval.evaluateClusterer(test);
    long testTimeElapsed = System.currentTimeMillis() - testTimeStart;
    //    m_result = eval.toSummaryString();


    // The results stored are all per instance -- can be multiplied by the
    // number of instances to get absolute numbers
    int current = 0;
    result[current++] = new Double(train.numInstances());
    result[current++] = new Double(test.numInstances());


    result[current++] = new Double(eval.getLogLikelihood());
    result[current++] = new Double(numClusters);
    
    // Timing stats
    result[current++] = new Double(trainTimeElapsed / 1000.0);
    result[current++] = new Double(testTimeElapsed / 1000.0);

View Full Code Here

      }
      // copy full model for output
      SerializedObject so = new SerializedObject(clusterer);
      fullClusterer = (Clusterer) so.getObject();
      
      ClusterEvaluation eval = new ClusterEvaluation();
      eval.setClusterer(clusterer);
      switch (testMode) {
        case 3: case 5: // Test on training
        m_Log.statusMessage("Clustering training data...");
        eval.evaluateClusterer(trainInst);
        predData = setUpVisualizableInstances(inst,eval);
        outBuff.append("=== Model and evaluation on training set ===\n\n");
        break;


        case 2: // Percent split
        m_Log.statusMessage("Randomizing instances...");
        inst.randomize(new Random(1));
        trainInst.randomize(new Random(1));
        int trainSize = trainInst.numInstances() * percent / 100;
        int testSize = trainInst.numInstances() - trainSize;
        Instances train = new Instances(trainInst, 0, trainSize);
        Instances test = new Instances(trainInst, trainSize, testSize);
        Instances testVis = new Instances(inst, trainSize, testSize);
        m_Log.statusMessage("Building model on training split...");
        clusterer.buildClusterer(train);
        m_Log.statusMessage("Evaluating on test split...");
        eval.evaluateClusterer(test);
        predData = setUpVisualizableInstances(testVis, eval);
        outBuff.append("=== Model and evaluation on test split ===\n");
        break;
    
        case 4: // Test on user split
        m_Log.statusMessage("Evaluating on test data...");
        Instances userTestT = new Instances(userTest);
        if (!m_ignoreKeyList.isSelectionEmpty()) {
    userTestT = removeIgnoreCols(userTestT);
        }
        eval.evaluateClusterer(userTestT);
        predData = setUpVisualizableInstances(userTest, eval);
        outBuff.append("=== Model and evaluation on test set ===\n");
        break;


        default:
        throw new Exception("Test mode not implemented");
      }
      outBuff.append(eval.clusterResultsToString());
      outBuff.append("\n");
      m_History.updateResult(name);
      m_Log.logMessage("Finished " + cname);
      m_Log.statusMessage("OK");
    } catch (Exception ex) {

View Full Code Here


              m_Log.logMessage("Started reevaluate model");
              if (m_Log instanceof TaskLogger) {
                ((TaskLogger)m_Log).taskStarted();
              }
              ClusterEvaluation eval = new ClusterEvaluation();
              eval.setClusterer(clusterer);
    
              Instances userTestT = new Instances(userTest);
              if (ignoredAtts != null) {
                userTestT = removeIgnoreCols(userTestT, ignoredAtts);
              }


              eval.evaluateClusterer(userTestT);
      
              predData = setUpVisualizableInstances(userTest, eval);


              outBuff.append("\n=== Re-evaluation on test set ===\n\n");
              outBuff.append("User supplied test set\n");  
              outBuff.append("Relation:     " + userTest.relationName() + '\n');
              outBuff.append("Instances:    " + userTest.numInstances() + '\n');
              outBuff.append("Attributes:   " + userTest.numAttributes() + "\n\n");
              if (trainHeader == null)
                outBuff.append("NOTE - if test set is not compatible then results are "
                               + "unpredictable\n\n");
      
              outBuff.append(eval.clusterResultsToString());
              outBuff.append("\n");
              m_History.updateResult(name);
              m_Log.logMessage("Finished re-evaluation");
              m_Log.statusMessage("OK");
            } catch (Exception ex) {

View Full Code Here

0 1

TOP

Related Classes of weka.clusterers.ClusterEvaluation

weka.classifiers.meta.ClassificationViaClustering

weka.core.converters.ConverterUtils.DataSource

weka.experiment.DensityBasedClustererSplitEvaluator

weka.filters.Filter

weka.filters.unsupervised.attribute.Remove

weka.gui.beans.ClustererPerformanceEvaluator

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.