Changeset 7234

Timestamp:

12/25/11 21:29:10 (13 years ago)

Author:

gkronber

Message:

#1685: changed simplification view for symbolic classification solutions to use the Gini index to determine the impact of a node it describes the degree of separation of the classes and we do not have to search for the optimal threshold value each time we calculate the impact of one node. Also fixed a problem with the Gini index result of classification solutions using a discriminating function as for these solutions the Gini index is calculated twice (once for the class values and once for the output values of the discriminating function.)

Location:

trunk/sources

Files:

• HeuristicLab.Problems.DataAnalysis.Symbolic.Classification.Views/3.4/InteractiveSymbolicDiscriminantFunctionClassificationSolutionSimplifierView.cs (modified) (2 diffs)
• HeuristicLab.Problems.DataAnalysis.Symbolic.Classification/3.4/SymbolicDiscriminantFunctionClassificationSolution.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/DiscriminantFunctionClassificationSolutionBase.cs (modified) (1 diff)

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic.Classification.Views/3.4/InteractiveSymbolicDiscriminantFunctionClassificationSolutionSimplifierView.cs

@@ -78 +78 @@
         .LimitToRange(Content.Model.LowerEstimationLimit, Content.Model.UpperEstimationLimit)
         .ToArray();
-      double[] classValues;
-      double[] thresholds;
-      // normal distribution cut points are used as thresholds here because they are a lot faster to calculate than the accuracy maximizing thresholds
-      AccuracyMaximizationThresholdCalculator.CalculateThresholds(Content.ProblemData, originalOutput, targetClassValues, out classValues, out thresholds);
-      var classifier = new SymbolicDiscriminantFunctionClassificationModel(tree, interpreter);
-      classifier.SetThresholdsAndClassValues(thresholds, classValues);
       OnlineCalculatorError errorState;
-      double originalAccuracy = OnlineAccuracyCalculator.Calculate(targetClassValues, classifier.GetEstimatedClassValues(dataset, rows), out errorState);
-      if (errorState != OnlineCalculatorError.None) originalAccuracy = 0.0;
+      double originalGini = NormalizedGiniCalculator.Calculate(targetClassValues, originalOutput, out errorState);
+      if (errorState != OnlineCalculatorError.None) originalGini = 0.0;
 
       foreach (ISymbolicExpressionTreeNode node in nodes) {
@@ -96 +90 @@
           .LimitToRange(Content.Model.LowerEstimationLimit, Content.Model.UpperEstimationLimit)
           .ToArray();
-        AccuracyMaximizationThresholdCalculator.CalculateThresholds(Content.ProblemData, newOutput, targetClassValues, out classValues, out thresholds);
-        classifier = new SymbolicDiscriminantFunctionClassificationModel(tree, interpreter);
-        classifier.SetThresholdsAndClassValues(thresholds, classValues);
-        double newAccuracy = OnlineAccuracyCalculator.Calculate(targetClassValues, classifier.GetEstimatedClassValues(dataset, rows), out errorState);
-        if (errorState != OnlineCalculatorError.None) newAccuracy = 0.0;
+        double newGini = NormalizedGiniCalculator.Calculate(targetClassValues, newOutput, out errorState);
+        if (errorState != OnlineCalculatorError.None) newGini = 0.0;
 
         // impact = 0 if no change
         // impact < 0 if new solution is better
         // impact > 0 if new solution is worse
-        impactValues[node] = originalAccuracy - newAccuracy;
+        impactValues[node] = originalGini - newGini;
         SwitchNode(parent, replacementNode, node);
       }

trunk/sources/HeuristicLab.Problems.DataAnalysis.Symbolic.Classification/3.4/SymbolicDiscriminantFunctionClassificationSolution.cs

@@ -79 +79 @@
       ModelDepth = Model.SymbolicExpressionTree.Depth;
     }
-
   }
 }

trunk/sources/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/DiscriminantFunctionClassificationSolutionBase.cs

@@ -157 +157 @@
     protected virtual void OnModelThresholdsChanged(EventArgs e) {
       CalculateResults();
+      CalculateRegressionResults();
     }