Changeset 13089

Timestamp:

10/29/15 19:48:17 (9 years ago)

Author:

gkronber

Message:

#1998

• deleted obsolete version of OneR algorithm (also does perform worse than mkommend's implementation in my tests)
• reused the ConstantRegressionModel as ConstantClassificationModel (OK?)
• fixed a few strings here and there

Location:

branches/ClassificationModelComparison

Files:

• HeuristicLab.Algorithms.DataAnalysis/3.4/HeuristicLab.Algorithms.DataAnalysis-3.4.csproj (modified) (1 diff)
• HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/OneR.cs (deleted)
• HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/OneRTest.cs (modified) (2 diffs)
• HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/ZeroR.cs (modified) (3 diffs)
• HeuristicLab.Problems.DataAnalysis.Views/3.4/Classification/ClassificationSolutionComparisonView.cs (modified) (2 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/HeuristicLab.Problems.DataAnalysis-3.4.csproj (modified) (1 diff)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ConstantClassificationModel.cs (deleted)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ConstantClassificationSolution.cs (modified) (2 diffs)
• HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/ConstantRegressionModel.cs (modified) (4 diffs)

branches/ClassificationModelComparison/HeuristicLab.Algorithms.DataAnalysis/3.4/HeuristicLab.Algorithms.DataAnalysis-3.4.csproj

@@ -318 +318 @@
     <Compile Include="Linear\AlglibUtil.cs" />
     <Compile Include="Linear\OneRTest.cs" />
-    <Compile Include="Linear\OneR.cs" />
     <Compile Include="Linear\OneR\OneRClassificationModel.cs" />
     <Compile Include="Linear\OneR\OneRClassificationSolution.cs" />

branches/ClassificationModelComparison/HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/OneRTest.cs

@@ -34 +34 @@
   /// 1R classification algorithm.
   /// </summary>
-  [Item("OneR Classification", "1R classification algorithm.")]
+  [Item("OneR Classification", "A simple classification algorithm the searches the best single-variable split (does not support categorical features correctly).")]
   [StorableClass]
   public sealed class OneRTest : FixedDataAnalysisAlgorithm<IClassificationProblem> {
@@ -63 +63 @@
     }
 
-    public static IClassificationSolution CreateOneRSolution(IClassificationProblemData problemData, int minBucketSize) {
+    public static IClassificationSolution CreateOneRSolution(IClassificationProblemData problemData, int minBucketSize = 6) {
       var bestClassified = 0;
       List<Split> bestSplits = null;

branches/ClassificationModelComparison/HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/ZeroR.cs

@@ -31 +31 @@
   /// 0R classification algorithm.
   /// </summary>
-  [Item("ZeroR", "0R classification algorithm.")]
-  [Creatable("Data Analysis")]
+  [Item("ZeroR", "The simplest possible classifier, ZeroR always predicts the majority class.")]
+  [Creatable("Data Analysis")] // TODO
   [StorableClass]
   public sealed class ZeroR : FixedDataAnalysisAlgorithm<IClassificationProblem> {
@@ -52 +52 @@
     protected override void Run() {
       var solution = CreateZeroRSolution(Problem.ProblemData);
-      Results.Add(new Result("ZeroR solution", "The 0R classifier.", solution));
+      Results.Add(new Result("ZeroR solution", "The simplest possible classifier, ZeroR always predicts the majority class.", solution));
     }
 
@@ -60 +60 @@
       var targetValues = dataset.GetDoubleValues(target, problemData.TrainingIndices);
 
+
+      // if multiple classes have the same number of observations then simply take the first one
       var dominantClass = targetValues.GroupBy(x => x).ToDictionary(g => g.Key, g => g.Count())
         .MaxItems(kvp => kvp.Value).Select(x => x.Key).First();
 
-      var model = new ConstantClassificationModel(dominantClass);
+      var model = new ConstantRegressionModel(dominantClass);
       var solution = new ConstantClassificationSolution(model, (IClassificationProblemData)problemData.Clone());
       return solution;

branches/ClassificationModelComparison/HeuristicLab.Problems.DataAnalysis.Views/3.4/Classification/ClassificationSolutionComparisonView.cs

@@ -87 +87 @@
           dataGridView.ColumnCount = 4;
           dataGridView.RowCount = solutions.Count();
-          dataGridView.Columns[0].HeaderText = "Training Accuracy";
-          dataGridView.Columns[1].HeaderText = "Test Accuracy";
-          dataGridView.Columns[2].HeaderText = "Matthews Correlation Coefficient Training";
-          dataGridView.Columns[3].HeaderText = "Matthews Correlation Coefficient Test";
+          dataGridView.Columns[0].HeaderText = "Accuracy (training)";
+          dataGridView.Columns[1].HeaderText = "Accuracy (test)";
+          dataGridView.Columns[2].HeaderText = "Matthews Correlation Coefficient (training)";
+          dataGridView.Columns[3].HeaderText = "Matthews Correlation Coefficient (test)";
           if (problemData.Classes == 2) {
             dataGridView.ColumnCount = 6;
-            dataGridView.Columns[4].HeaderText = "F1 Score Training";
-            dataGridView.Columns[5].HeaderText = "F1 Score Test";
+            dataGridView.Columns[4].HeaderText = "F1 Score (training)";
+            dataGridView.Columns[5].HeaderText = "F1 Score (test)";
           }
 
@@ -130 +130 @@
       var newSolutions = new List<IClassificationSolution>();
       var zeroR = ZeroR.CreateZeroRSolution(problemData);
-      zeroR.Name = "0R Classification Solution";
+      zeroR.Name = "ZeroR Classification Solution";
       newSolutions.Add(zeroR);
-      var oneR = OneR.CreateOneRSolution(problemData, 6, new FastRandom());
-      oneR.Name = "1R Classification Solution";
+      var oneR = OneRTest.CreateOneRSolution(problemData);
+      oneR.Name = "OneR Classification Solution";
       newSolutions.Add(oneR);
       try {

branches/ClassificationModelComparison/HeuristicLab.Problems.DataAnalysis/3.4/HeuristicLab.Problems.DataAnalysis-3.4.csproj

@@ -172 +172 @@
     <Compile Include="Implementation\Classification\ConstantClassificationSolution.cs" />
     <Compile Include="Implementation\Classification\DiscriminantFunctionClassificationSolutionBase.cs" />
-    <Compile Include="Implementation\Classification\ConstantClassificationModel.cs" />
     <Compile Include="Implementation\Clustering\ClusteringProblem.cs" />
     <Compile Include="Implementation\Clustering\ClusteringProblemData.cs" />

branches/ClassificationModelComparison/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ConstantClassificationSolution.cs

@@ -28 +28 @@
   [Item(Name = "Constant Classification Solution", Description = "Represents a constant classification solution (model + data).")]
   public class ConstantClassificationSolution : ClassificationSolution {
-    public new ConstantClassificationModel Model {
-      get { return (ConstantClassificationModel)base.Model; }
+    public new ConstantRegressionModel Model {
+      get { return (ConstantRegressionModel)base.Model; }
       set { base.Model = value; }
     }
@@ -36 +36 @@
     protected ConstantClassificationSolution(bool deserializing) : base(deserializing) { }
     protected ConstantClassificationSolution(ConstantClassificationSolution original, Cloner cloner) : base(original, cloner) { }
-    public ConstantClassificationSolution(ConstantClassificationModel model, IClassificationProblemData problemData)
+    public ConstantClassificationSolution(ConstantRegressionModel model, IClassificationProblemData problemData)
       : base(model, problemData) {
       RecalculateResults();

branches/ClassificationModelComparison/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Regression/ConstantRegressionModel.cs

@@ -30 +30 @@
 namespace HeuristicLab.Problems.DataAnalysis {
   [StorableClass]
-  [Item("Constant Regression Model", "A model that always returns the same constant value regardless of the presented input data.")]
-  public class ConstantRegressionModel : NamedItem, IRegressionModel, IStringConvertibleValue {
+  [Item("Constant Model", "A model that always returns the same constant value regardless of the presented input data.")]
+  // TODO this class should be renamed, since it is also used to handle the classification case now
+  public class ConstantRegressionModel : NamedItem, IRegressionModel, IClassificationModel, IStringConvertibleValue {
     [Storable]
     private double constant;
@@ -59 +60 @@
     }
 
+    public IEnumerable<double> GetEstimatedClassValues(IDataset dataset, IEnumerable<int> rows) {
+      return GetEstimatedValues(dataset, rows);
+    }
+
     public IRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
       return new ConstantRegressionSolution(this, new RegressionProblemData(problemData));
+    }
+
+    public IClassificationSolution CreateClassificationSolution(IClassificationProblemData problemData) {
+      return new ConstantClassificationSolution(this, new ClassificationProblemData(problemData));
     }
 
@@ -70 +79 @@
     public bool ReadOnly { get; private set; }
     public bool Validate(string value, out string errorMessage) {
-      throw new NotSupportedException(); // changing a constant regression model is not supported
+      throw new NotSupportedException(); // changing a constant model is not supported
     }
 
@@ -78 +87 @@
 
     public bool SetValue(string value) {
-      throw new NotSupportedException(); // changing a constant regression model is not supported
+      throw new NotSupportedException(); // changing a constant model is not supported
     }
 
     public event EventHandler ValueChanged;
     #endregion
+
   }
 }