Changeset 17808 for stable/HeuristicLab.Problems.DataAnalysis.Views/3.4/Regression

Timestamp:

12/30/20 09:47:47 (4 years ago)

Author:

gkronber

Message:

#3038: merged r17341 and r17539 from trunk to stable

Location:

stable

Files:

• . (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis.Views (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis.Views/3.4 (modified) (1 prop)
• HeuristicLab.Problems.DataAnalysis.Views/3.4/Regression/RegressionSolutionResidualAnalysisView.cs (modified) (3 diffs)

stable/HeuristicLab.Problems.DataAnalysis.Views/3.4/Regression/RegressionSolutionResidualAnalysisView.cs

@@ -88 +88 @@
       var ds = problemData.Dataset;
       var runs = new RunCollection();
-      // determine relevant variables (at least two different values)
-      var doubleVars = ds.DoubleVariables.Where(vn => ds.GetDoubleValues(vn).Distinct().Skip(1).Any()).ToArray();
-      var stringVars = ds.StringVariables.Where(vn => ds.GetStringValues(vn).Distinct().Skip(1).Any()).ToArray();
-      var dateTimeVars = ds.DateTimeVariables.Where(vn => ds.GetDateTimeValues(vn).Distinct().Skip(1).Any()).ToArray();
 
-      // produce training and test values separately as they might overlap (e.g. for ensembles)
-      var predictedValuesTrain = Content.EstimatedTrainingValues.ToArray();
-      int j = 0; // idx for predictedValues array
-      foreach (var i in problemData.TrainingIndices) {
+      var doubleVars = ds.DoubleVariables.ToArray();
+      var stringVars = ds.StringVariables.ToArray();
+      var dateTimeVars = ds.DateTimeVariables.ToArray();
+
+      var predictedValues = Content.EstimatedValues.ToArray();
+      var targetValues = ds.GetReadOnlyDoubleValues(problemData.TargetVariable);
+
+      foreach (var i in problemData.AllIndices) {
         var run = CreateRunForIdx(i, problemData, doubleVars, stringVars, dateTimeVars);
-        var targetValue = ds.GetDoubleValue(problemData.TargetVariable, i);
-        AddErrors(run, predictedValuesTrain[j++], targetValue);
-        run.Results.Add(PartitionLabel, new StringValue("Training"));
-        run.Color = Color.Gold;
+        AddErrors(run, predictedValues[i], targetValues[i]);
+
+        if (problemData.IsTrainingSample(i) && problemData.IsTestSample(i)) {
+          run.Results.Add(PartitionLabel, new StringValue("Training + Test"));
+          run.Color = Color.Orange;
+        } else if (problemData.IsTrainingSample(i)) {
+          run.Results.Add(PartitionLabel, new StringValue("Training"));
+          run.Color = Color.Gold;
+        } else if (problemData.IsTestSample(i)) {
+          run.Results.Add(PartitionLabel, new StringValue("Test"));
+          run.Color = Color.Red;
+        } else {
+          run.Results.Add(PartitionLabel, new StringValue("Additional Data"));
+          run.Color = Color.Black;
+        }
         runs.Add(run);
       }
-      var predictedValuesTest = Content.EstimatedTestValues.ToArray();
-      j = 0;
-      foreach (var i in problemData.TestIndices) {
-        var run = CreateRunForIdx(i, problemData, doubleVars, stringVars, dateTimeVars);
-        var targetValue = ds.GetDoubleValue(problemData.TargetVariable, i);
-        AddErrors(run, predictedValuesTest[j++], targetValue);
-        run.Results.Add(PartitionLabel, new StringValue("Test"));
-        run.Color = Color.Red;
-        runs.Add(run);
-      }
+
       if (string.IsNullOrEmpty(selectedXAxis))
         selectedXAxis = "Index";
@@ -124 +126 @@
     }
 
-    private void AddErrors(IRun run, double pred, double target) {
+    private static void AddErrors(IRun run, double pred, double target) {
       var residual = target - pred;
       var relError = residual / target;
@@ -135 +137 @@
     }
 
-    private IRun CreateRunForIdx(int i, IRegressionProblemData problemData, IEnumerable<string> doubleVars, IEnumerable<string> stringVars, IEnumerable<string> dateTimeVars) {
+    private static IRun CreateRunForIdx(int i, IRegressionProblemData problemData, IEnumerable<string> doubleVars, IEnumerable<string> stringVars, IEnumerable<string> dateTimeVars) {
       var ds = problemData.Dataset;
       var run = new Run();