Free cookie consent management tool by TermsFeed Policy Generator

source: branches/2892_LR-prediction-intervals/HeuristicLab.Problems.DataAnalysis/3.4/Implementation/Classification/ClassificationSolutionBase.cs @ 16388

Last change on this file since 16388 was 16388, checked in by gkronber, 5 years ago

#2892: Merging r15750:16382 (HEAD) from trunk to branch, resolving conflicts

File size: 10.4 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2018 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
4 *
5 * This file is part of HeuristicLab.
6 *
7 * HeuristicLab is free software: you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation, either version 3 of the License, or
10 * (at your option) any later version.
11 *
12 * HeuristicLab is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
15 * GNU General Public License for more details.
16 *
17 * You should have received a copy of the GNU General Public License
18 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
19 */
20#endregion
21
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using HeuristicLab.Common;
26using HeuristicLab.Data;
27using HeuristicLab.Optimization;
28using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
29using HeuristicLab.Problems.DataAnalysis.OnlineCalculators;
30
31namespace HeuristicLab.Problems.DataAnalysis {
32  [StorableClass]
33  public abstract class ClassificationSolutionBase : DataAnalysisSolution, IClassificationSolution {
34    private const string TrainingAccuracyResultName = "Accuracy (training)";
35    private const string TestAccuracyResultName = "Accuracy (test)";
36    private const string TrainingNormalizedGiniCoefficientResultName = "Normalized Gini Coefficient (training)";
37    private const string TestNormalizedGiniCoefficientResultName = "Normalized Gini Coefficient (test)";
38    private const string ClassificationPerformanceMeasuresResultName = "Classification Performance Measures";
39
40    public new IClassificationModel Model {
41      get { return (IClassificationModel)base.Model; }
42      protected set { base.Model = value; }
43    }
44
45    public new IClassificationProblemData ProblemData {
46      get { return (IClassificationProblemData)base.ProblemData; }
47      set {
48        if (value == null) throw new ArgumentNullException("The problemData must not be null.");
49        string errorMessage = string.Empty;
50        if (!Model.IsProblemDataCompatible(value, out errorMessage)) throw new ArgumentException(errorMessage);
51
52        base.ProblemData = value;
53      }
54    }
55
56    #region Results
57    public double TrainingAccuracy {
58      get { return ((DoubleValue)this[TrainingAccuracyResultName].Value).Value; }
59      private set { ((DoubleValue)this[TrainingAccuracyResultName].Value).Value = value; }
60    }
61    public double TestAccuracy {
62      get { return ((DoubleValue)this[TestAccuracyResultName].Value).Value; }
63      private set { ((DoubleValue)this[TestAccuracyResultName].Value).Value = value; }
64    }
65    public double TrainingNormalizedGiniCoefficient {
66      get { return ((DoubleValue)this[TrainingNormalizedGiniCoefficientResultName].Value).Value; }
67      protected set { ((DoubleValue)this[TrainingNormalizedGiniCoefficientResultName].Value).Value = value; }
68    }
69    public double TestNormalizedGiniCoefficient {
70      get { return ((DoubleValue)this[TestNormalizedGiniCoefficientResultName].Value).Value; }
71      protected set { ((DoubleValue)this[TestNormalizedGiniCoefficientResultName].Value).Value = value; }
72    }
73    public ClassificationPerformanceMeasuresResultCollection ClassificationPerformanceMeasures {
74      get { return ((ClassificationPerformanceMeasuresResultCollection)this[ClassificationPerformanceMeasuresResultName].Value); }
75      protected set { (this[ClassificationPerformanceMeasuresResultName].Value) = value; }
76    }
77    #endregion
78
79    [StorableConstructor]
80    protected ClassificationSolutionBase(bool deserializing) : base(deserializing) { }
81    protected ClassificationSolutionBase(ClassificationSolutionBase original, Cloner cloner)
82      : base(original, cloner) {
83    }
84    protected ClassificationSolutionBase(IClassificationModel model, IClassificationProblemData problemData)
85      : base(model, problemData) {
86      Add(new Result(TrainingAccuracyResultName, "Accuracy of the model on the training partition (percentage of correctly classified instances).", new PercentValue()));
87      Add(new Result(TestAccuracyResultName, "Accuracy of the model on the test partition (percentage of correctly classified instances).", new PercentValue()));
88      Add(new Result(TrainingNormalizedGiniCoefficientResultName, "Normalized Gini coefficient of the model on the training partition.", new DoubleValue()));
89      Add(new Result(TestNormalizedGiniCoefficientResultName, "Normalized Gini coefficient of the model on the test partition.", new DoubleValue()));
90      Add(new Result(ClassificationPerformanceMeasuresResultName, @"Classification performance measures.\n
91                              In a multiclass classification all misclassifications of the negative class will be treated as true negatives except on positive class estimations.",
92                            new ClassificationPerformanceMeasuresResultCollection()));
93    }
94
95    [StorableHook(HookType.AfterDeserialization)]
96    private void AfterDeserialization() {
97      if (string.IsNullOrEmpty(Model.TargetVariable))
98        Model.TargetVariable = this.ProblemData.TargetVariable;
99
100      if (!this.ContainsKey(TrainingNormalizedGiniCoefficientResultName))
101        Add(new Result(TrainingNormalizedGiniCoefficientResultName, "Normalized Gini coefficient of the model on the training partition.", new DoubleValue()));
102      if (!this.ContainsKey(TestNormalizedGiniCoefficientResultName))
103        Add(new Result(TestNormalizedGiniCoefficientResultName, "Normalized Gini coefficient of the model on the test partition.", new DoubleValue()));
104      if (!this.ContainsKey(ClassificationPerformanceMeasuresResultName)) {
105        Add(new Result(ClassificationPerformanceMeasuresResultName, @"Classification performance measures.\n
106                              In a multiclass classification all misclassifications of the negative class will be treated as true negatives except on positive class estimations.",
107                              new ClassificationPerformanceMeasuresResultCollection()));
108        CalculateClassificationResults();
109      }
110    }
111
112    protected void CalculateClassificationResults() {
113      double[] estimatedTrainingClassValues = EstimatedTrainingClassValues.ToArray(); // cache values
114      double[] originalTrainingClassValues = ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TrainingIndices).ToArray();
115
116      double[] estimatedTestClassValues = EstimatedTestClassValues.ToArray(); // cache values
117      double[] originalTestClassValues = ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TestIndices).ToArray();
118
119      var positiveClassName = ProblemData.PositiveClass;
120      double positiveClassValue = ProblemData.GetClassValue(positiveClassName);
121      ClassificationPerformanceMeasuresCalculator trainingPerformanceCalculator = new ClassificationPerformanceMeasuresCalculator(positiveClassName, positiveClassValue);
122      ClassificationPerformanceMeasuresCalculator testPerformanceCalculator = new ClassificationPerformanceMeasuresCalculator(positiveClassName, positiveClassValue);
123
124      OnlineCalculatorError errorState;
125      double trainingAccuracy = OnlineAccuracyCalculator.Calculate(originalTrainingClassValues, estimatedTrainingClassValues, out errorState);
126      if (errorState != OnlineCalculatorError.None) trainingAccuracy = double.NaN;
127      double testAccuracy = OnlineAccuracyCalculator.Calculate(originalTestClassValues, estimatedTestClassValues, out errorState);
128      if (errorState != OnlineCalculatorError.None) testAccuracy = double.NaN;
129
130      TrainingAccuracy = trainingAccuracy;
131      TestAccuracy = testAccuracy;
132
133      double trainingNormalizedGini = NormalizedGiniCalculator.Calculate(originalTrainingClassValues, estimatedTrainingClassValues, out errorState);
134      if (errorState != OnlineCalculatorError.None) trainingNormalizedGini = double.NaN;
135      double testNormalizedGini = NormalizedGiniCalculator.Calculate(originalTestClassValues, estimatedTestClassValues, out errorState);
136      if (errorState != OnlineCalculatorError.None) testNormalizedGini = double.NaN;
137
138      TrainingNormalizedGiniCoefficient = trainingNormalizedGini;
139      TestNormalizedGiniCoefficient = testNormalizedGini;
140
141      ClassificationPerformanceMeasures.Reset();
142
143      trainingPerformanceCalculator.Calculate(originalTrainingClassValues, estimatedTrainingClassValues);
144      if (trainingPerformanceCalculator.ErrorState == OnlineCalculatorError.None)
145        ClassificationPerformanceMeasures.SetTrainingResults(trainingPerformanceCalculator);
146
147      testPerformanceCalculator.Calculate(originalTestClassValues, estimatedTestClassValues);
148      if (testPerformanceCalculator.ErrorState == OnlineCalculatorError.None)
149        ClassificationPerformanceMeasures.SetTestResults(testPerformanceCalculator);
150
151      if (ProblemData.Classes == 2) {
152        var f1Training = FOneScoreCalculator.Calculate(originalTrainingClassValues, estimatedTrainingClassValues, out errorState);
153        if (errorState == OnlineCalculatorError.None) ClassificationPerformanceMeasures.TrainingF1Score = f1Training;
154        var f1Test = FOneScoreCalculator.Calculate(originalTestClassValues, estimatedTestClassValues, out errorState);
155        if (errorState == OnlineCalculatorError.None) ClassificationPerformanceMeasures.TestF1Score = f1Test;
156      }
157
158      var mccTraining = MatthewsCorrelationCoefficientCalculator.Calculate(originalTrainingClassValues, estimatedTrainingClassValues, out errorState);
159      if (errorState == OnlineCalculatorError.None) ClassificationPerformanceMeasures.TrainingMatthewsCorrelation = mccTraining;
160      var mccTest = MatthewsCorrelationCoefficientCalculator.Calculate(originalTestClassValues, estimatedTestClassValues, out errorState);
161      if (errorState == OnlineCalculatorError.None) ClassificationPerformanceMeasures.TestMatthewsCorrelation = mccTest;
162    }
163
164    public abstract IEnumerable<double> EstimatedClassValues { get; }
165    public abstract IEnumerable<double> EstimatedTrainingClassValues { get; }
166    public abstract IEnumerable<double> EstimatedTestClassValues { get; }
167
168    public abstract IEnumerable<double> GetEstimatedClassValues(IEnumerable<int> rows);
169
170    protected override void RecalculateResults() {
171      CalculateClassificationResults();
172    }
173  }
174}
Note: See TracBrowser for help on using the repository browser.