Changeset 8114

Timestamp:

06/26/12 09:15:55 (12 years ago)

Author:

mkommend

Message:

#1081: Corrected scaling behavior of symbolic time series models.

Location:

branches/HeuristicLab.TimeSeries

Files:

• HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis.Views/3.4/InteractiveSymbolicTimeSeriesPrognosisSolutionSimplifierView.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveMeanSquaredErrorEvaluator.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveTrainingBestSolutionAnalyzer.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveValidationBestSolutionAnalyzer.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SymbolicTimeSeriesPrognosisModel.cs (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Grammars/TypeCoherentExpressionGrammar.cs (modified) (1 diff)

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis.Views/3.4/InteractiveSymbolicTimeSeriesPrognosisSolutionSimplifierView.cs

@@ -51 +51 @@
     protected override void UpdateModel(ISymbolicExpressionTree tree) {
       var model = new SymbolicTimeSeriesPrognosisModel(tree, Content.Model.Interpreter);
-      SymbolicTimeSeriesPrognosisModel.Scale(model, Content.ProblemData);
+      SymbolicTimeSeriesPrognosisModel.Scale(model, Content.ProblemData, Content.ProblemData.TrainingIndizes);
       Content.Model = model;
     }

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveMeanSquaredErrorEvaluator.cs

@@ -69 +69 @@
 
       double mse;
-      if (applyLinearScaling) {
+      if (applyLinearScaling && horizon == 1) { //perform normal evaluation and afterwards scale the solution and calculate the fitness value        
         var mseCalculator = new OnlineMeanSquaredErrorCalculator();
         CalculateWithScaling(targetValues, boundedEstimatedValues, mseCalculator, problemData.Dataset.Rows * horizon);
         errorState = mseCalculator.ErrorState;
         mse = mseCalculator.MeanSquaredError;
+      } else if (applyLinearScaling) { //first create model to perform linear scaling and afterwards calculate fitness for the scaled model
+        var model = new SymbolicTimeSeriesPrognosisModel((ISymbolicExpressionTree)solution.Clone(), interpreter, lowerEstimationLimit, upperEstimationLimit);
+        SymbolicTimeSeriesPrognosisModel.Scale(model, problemData, rows);
+        var scaledSolution = model.SymbolicExpressionTree;
+        estimatedValues = interpreter.GetSymbolicExpressionTreeValues(scaledSolution, problemData.Dataset, rows, horizions).SelectMany(x => x);
+        boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
+        mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);
       } else
         mse = OnlineMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, out errorState);

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveTrainingBestSolutionAnalyzer.cs

@@ -77 +77 @@
       var model = new SymbolicTimeSeriesPrognosisModel((ISymbolicExpressionTree)bestTree.Clone(), SymbolicDataAnalysisTreeInterpreterParameter.ActualValue as ISymbolicTimeSeriesPrognosisExpressionTreeInterpreter, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper);
       if (ApplyLinearScaling.Value)
-        SymbolicTimeSeriesPrognosisModel.Scale(model, ProblemDataParameter.ActualValue);
+        SymbolicTimeSeriesPrognosisModel.Scale(model, ProblemDataParameter.ActualValue, ProblemDataParameter.ActualValue.TrainingIndizes);
       return new SymbolicTimeSeriesPrognosisSolution(model, (ITimeSeriesPrognosisProblemData)ProblemDataParameter.ActualValue.Clone());
     }

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SingleObjective/SymbolicTimeSeriesPrognosisSingleObjectiveValidationBestSolutionAnalyzer.cs

@@ -66 +66 @@
       var model = new SymbolicTimeSeriesPrognosisModel((ISymbolicExpressionTree)bestTree.Clone(), SymbolicDataAnalysisTreeInterpreterParameter.ActualValue as ISymbolicTimeSeriesPrognosisExpressionTreeInterpreter, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper);
       if (ApplyLinearScaling.Value)
-        SymbolicTimeSeriesPrognosisModel.Scale(model, ProblemDataParameter.ActualValue);
+        SymbolicTimeSeriesPrognosisModel.Scale(model, ProblemDataParameter.ActualValue, ProblemDataParameter.ActualValue.TrainingIndizes);
       return new SymbolicTimeSeriesPrognosisSolution(model, (ITimeSeriesPrognosisProblemData)ProblemDataParameter.ActualValue.Clone());
     }

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic.TimeSeriesPrognosis/3.4/SymbolicTimeSeriesPrognosisModel.cs

@@ -98 +98 @@
     }
 
-    public static void Scale(SymbolicTimeSeriesPrognosisModel model, ITimeSeriesPrognosisProblemData problemData) {
+    public static void Scale(SymbolicTimeSeriesPrognosisModel model, ITimeSeriesPrognosisProblemData problemData, IEnumerable<int> rows) {
       var dataset = problemData.Dataset;
       var targetVariable = problemData.TargetVariable;
-      var rows = problemData.TrainingIndizes;
-      var estimatedValuesEnumerator = model.Interpreter.GetSymbolicExpressionTreeValues(model.SymbolicExpressionTree, dataset, rows);
-      var targetValuesEnumerator = problemData.Dataset.GetDoubleValues(targetVariable, rows);
+      var estimatedValues = model.Interpreter.GetSymbolicExpressionTreeValues(model.SymbolicExpressionTree, dataset, rows);
+      var boundedEstimatedValues = estimatedValues.LimitToRange(model.lowerEstimationLimit, model.upperEstimationLimit);
+      var targetValues = problemData.Dataset.GetDoubleValues(targetVariable, rows);
 
       double alpha, beta;
       OnlineCalculatorError error;
-      OnlineLinearScalingParameterCalculator.Calculate(estimatedValuesEnumerator, targetValuesEnumerator, out alpha, out beta, out error);
+      OnlineLinearScalingParameterCalculator.Calculate(boundedEstimatedValues, targetValues, out alpha, out beta, out error);
       if (error != OnlineCalculatorError.None) return;
 

branches/HeuristicLab.TimeSeries/HeuristicLab.Problems.DataAnalysis.Symbolic/3.4/Grammars/TypeCoherentExpressionGrammar.cs

@@ -227 +227 @@
 
     public void ConfigureAsDefaultTimeSeriesPrognosisGrammar() {
-      Symbols.Where(s => s is Variable).First().Enabled = false;
-      Symbols.Where(s => s.Name == TrigonometricFunctionsName).First().Enabled = false;
-      Symbols.Where(s => s.Name == PowerFunctionsName).First().Enabled = false;
-      Symbols.Where(s => s.Name == ConditionalSymbolsName).First().Enabled = false;
+      Symbols.First(s => s is Variable).Enabled = false;
+      Symbols.First(s => s.Name == TrigonometricFunctionsName).Enabled = false;
+      Symbols.First(s => s.Name == PowerFunctionsName).Enabled = false;
+      Symbols.First(s => s.Name == ConditionalSymbolsName).Enabled = false;
+      Symbols.First(s => s.Name == SpecialFunctionsName).Enabled = false;
     }
   }