Changeset 10982

Timestamp:

06/11/14 13:38:58 (10 years ago)

Author:

rstoll

Message:

• ProblemCreator did not take into account:

non double input variables
treated target variable as an additional input variable
did not consider constant input variables

Location:

branches/DataPreprocessing/HeuristicLab.DataPreprocessing/3.4

Files:

• Implementations/CorrelationMatrixContent.cs (modified) (1 diff)
• ProblemDataCreator.cs (modified) (5 diffs)

branches/DataPreprocessing/HeuristicLab.DataPreprocessing/3.4/Implementations/CorrelationMatrixContent.cs

@@ -51 +51 @@
       : base(original, cloner) {
     }
+
     public override IDeepCloneable Clone(Cloner cloner) {
       return new CorrelationMatrixContent(this, cloner);

branches/DataPreprocessing/HeuristicLab.DataPreprocessing/3.4/ProblemDataCreator.cs

@@ -22 +22 @@
 using System;
 using System.Collections.Generic;
+using HeuristicLab.Common;
 using HeuristicLab.Problems.DataAnalysis;
 using HeuristicLab.Problems.DataAnalysis.Transformations;
@@ -35 +36 @@
     private Dataset exporteDataset;
 
-    private IEnumerable<string> InputVariables { get { return context.Data.VariableNames; } }
     private IList<ITransformation> Transformations { get { return context.Data.Transformations; } }
 
@@ -65 +65 @@
       var targetVariable = oldProblemData.TargetVariable;
       // target variable must be double and must exist in the new dataset
-      return new RegressionProblemData(ExportedDataset, InputVariables, targetVariable, Transformations);
+      return new RegressionProblemData(ExportedDataset, GetDoubleInputVariables(targetVariable), targetVariable, Transformations);
     }
 
@@ -71 +71 @@
       var targetVariable = oldProblemData.TargetVariable;
       // target variable must be double and must exist in the new dataset
-      return new ClassificationProblemData(ExportedDataset, InputVariables, targetVariable, Transformations);
+      return new ClassificationProblemData(ExportedDataset, GetDoubleInputVariables(targetVariable), targetVariable, Transformations);
     }
 
     private IDataAnalysisProblemData CreateClusteringData(ClusteringProblemData oldProblemData) {
-      return new ClusteringProblemData(ExportedDataset, InputVariables, Transformations);
+      return new ClusteringProblemData(ExportedDataset, GetDoubleInputVariables(String.Empty), Transformations);
     }
 
@@ -86 +86 @@
       problemData.TestPartition.End = ppData.TestPartition.End;
     }
+
+    private IEnumerable<string> GetDoubleInputVariables(string targetVariable) {
+      var variableNames = new List<string>();
+      for (int i = 0; i < context.Data.Columns; ++i) {
+        var variableName = context.Data.GetVariableName(i);
+        if (context.Data.IsType<double>(i)
+          && variableName != targetVariable
+          && IsNotConstantInputVariable(context.Data.GetValues<double>(i))) {
+
+          variableNames.Add(variableName);
+        }
+      }
+      return variableNames;
+    }
+
+    private bool IsNotConstantInputVariable(IList<double> list) {
+      return context.Data.TrainingPartition.End - context.Data.TrainingPartition.Start > 1 || list.Range() > 0;
+    }
   }
 }