[4366] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[5445] | 3 | * Copyright (C) 2002-2011 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[4366] | 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 |
|
---|
| 22 | using System;
|
---|
| 23 | using System.Collections.Generic;
|
---|
| 24 | using System.Linq;
|
---|
| 25 | using HeuristicLab.Common;
|
---|
| 26 | using HeuristicLab.Core;
|
---|
| 27 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 28 | using HeuristicLab.Problems.DataAnalysis.Regression.Symbolic;
|
---|
| 29 |
|
---|
| 30 | namespace HeuristicLab.Problems.DataAnalysis.Classification {
|
---|
| 31 | /// <summary>
|
---|
[4391] | 32 | /// Represents a solution for a symbolic classification problem which can be visualized in the GUI.
|
---|
[4366] | 33 | /// </summary>
|
---|
| 34 | [Item("SymbolicClassificationSolution", "Represents a solution for a symbolic classification problem which can be visualized in the GUI.")]
|
---|
| 35 | [StorableClass]
|
---|
[4417] | 36 | public class SymbolicClassificationSolution : SymbolicRegressionSolution, IClassificationSolution {
|
---|
[4722] | 37 | public new ClassificationProblemData ProblemData {
|
---|
| 38 | get { return (ClassificationProblemData)base.ProblemData; }
|
---|
| 39 | set { base.ProblemData = value; }
|
---|
| 40 | }
|
---|
| 41 |
|
---|
| 42 | #region properties
|
---|
| 43 | private List<double> optimalThresholds;
|
---|
| 44 | private List<double> actualThresholds;
|
---|
| 45 | public IEnumerable<double> Thresholds {
|
---|
| 46 | get {
|
---|
| 47 | if (actualThresholds == null) RecalculateEstimatedValues();
|
---|
| 48 | return actualThresholds;
|
---|
| 49 | }
|
---|
| 50 | set {
|
---|
| 51 | if (actualThresholds != null && actualThresholds.SequenceEqual(value))
|
---|
| 52 | return;
|
---|
| 53 | actualThresholds = new List<double>(value);
|
---|
| 54 | OnThresholdsChanged();
|
---|
| 55 | }
|
---|
| 56 | }
|
---|
| 57 |
|
---|
| 58 | public IEnumerable<double> EstimatedClassValues {
|
---|
| 59 | get { return GetEstimatedClassValues(Enumerable.Range(0, ProblemData.Dataset.Rows)); }
|
---|
| 60 | }
|
---|
| 61 |
|
---|
| 62 | public IEnumerable<double> EstimatedTrainingClassValues {
|
---|
| 63 | get { return GetEstimatedClassValues(ProblemData.TrainingIndizes); }
|
---|
| 64 | }
|
---|
| 65 |
|
---|
| 66 | public IEnumerable<double> EstimatedTestClassValues {
|
---|
| 67 | get { return GetEstimatedClassValues(ProblemData.TestIndizes); }
|
---|
| 68 | }
|
---|
| 69 |
|
---|
| 70 | [StorableConstructor]
|
---|
| 71 | protected SymbolicClassificationSolution(bool deserializing) : base(deserializing) { }
|
---|
| 72 | protected SymbolicClassificationSolution(SymbolicClassificationSolution original, Cloner cloner) : base(original, cloner) { }
|
---|
[4366] | 73 | public SymbolicClassificationSolution(ClassificationProblemData problemData, SymbolicRegressionModel model, double lowerEstimationLimit, double upperEstimationLimit)
|
---|
[4417] | 74 | : base(problemData, model, lowerEstimationLimit, upperEstimationLimit) {
|
---|
[4366] | 75 | }
|
---|
| 76 |
|
---|
[4722] | 77 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 78 | return new SymbolicClassificationSolution(this, cloner);
|
---|
[4366] | 79 | }
|
---|
| 80 |
|
---|
| 81 | protected override void RecalculateEstimatedValues() {
|
---|
| 82 | estimatedValues =
|
---|
| 83 | (from x in Model.GetEstimatedValues(ProblemData, 0, ProblemData.Dataset.Rows)
|
---|
| 84 | let boundedX = Math.Min(UpperEstimationLimit, Math.Max(LowerEstimationLimit, x))
|
---|
| 85 | select double.IsNaN(boundedX) ? UpperEstimationLimit : boundedX).ToList();
|
---|
| 86 | RecalculateClassIntermediates();
|
---|
[5323] | 87 | OnEstimatedValuesChanged();
|
---|
[4366] | 88 | }
|
---|
| 89 |
|
---|
| 90 | private void RecalculateClassIntermediates() {
|
---|
[4391] | 91 | int slices = 100;
|
---|
[4366] | 92 |
|
---|
[4391] | 93 | List<int> classInstances = (from classValue in ProblemData.Dataset.GetVariableValues(ProblemData.TargetVariable.Value)
|
---|
| 94 | group classValue by classValue into grouping
|
---|
| 95 | select grouping.Count()).ToList();
|
---|
| 96 |
|
---|
[4366] | 97 | List<KeyValuePair<double, double>> estimatedTargetValues =
|
---|
[4469] | 98 | (from row in ProblemData.TrainingIndizes
|
---|
[4391] | 99 | select new KeyValuePair<double, double>(
|
---|
| 100 | estimatedValues[row],
|
---|
| 101 | ProblemData.Dataset[ProblemData.TargetVariable.Value, row])).ToList();
|
---|
[4366] | 102 |
|
---|
[4391] | 103 | List<double> originalClasses = ProblemData.SortedClassValues.ToList();
|
---|
| 104 | double[] thresholds = new double[ProblemData.NumberOfClasses + 1];
|
---|
[4366] | 105 | thresholds[0] = double.NegativeInfinity;
|
---|
| 106 | thresholds[thresholds.Length - 1] = double.PositiveInfinity;
|
---|
| 107 |
|
---|
| 108 | for (int i = 1; i < thresholds.Length - 1; i++) {
|
---|
| 109 | double lowerThreshold = thresholds[i - 1];
|
---|
| 110 | double actualThreshold = originalClasses[i - 1];
|
---|
| 111 | double thresholdIncrement = (originalClasses[i] - originalClasses[i - 1]) / slices;
|
---|
| 112 |
|
---|
[5332] | 113 | double lowestBestThreshold = double.NaN;
|
---|
| 114 | double highestBestThreshold = double.NaN;
|
---|
[4391] | 115 | double bestClassificationScore = double.PositiveInfinity;
|
---|
[5341] | 116 | bool seriesOfEqualClassificationScores = false;
|
---|
[4366] | 117 |
|
---|
| 118 | while (actualThreshold < originalClasses[i]) {
|
---|
[4391] | 119 | double classificationScore = 0.0;
|
---|
[4366] | 120 |
|
---|
| 121 | foreach (KeyValuePair<double, double> estimatedTarget in estimatedTargetValues) {
|
---|
| 122 | //all positives
|
---|
| 123 | if (estimatedTarget.Value.IsAlmost(originalClasses[i - 1])) {
|
---|
| 124 | if (estimatedTarget.Key > lowerThreshold && estimatedTarget.Key < actualThreshold)
|
---|
[4391] | 125 | //true positive
|
---|
[5319] | 126 | classificationScore += ProblemData.MisclassificationMatrix[i - 1, i - 1];
|
---|
[4366] | 127 | else
|
---|
[4391] | 128 | //false negative
|
---|
[5319] | 129 | classificationScore += ProblemData.MisclassificationMatrix[i, i - 1];
|
---|
[4366] | 130 | }
|
---|
| 131 | //all negatives
|
---|
| 132 | else {
|
---|
| 133 | if (estimatedTarget.Key > lowerThreshold && estimatedTarget.Key < actualThreshold)
|
---|
[5332] | 134 | //false positive
|
---|
[5319] | 135 | classificationScore += ProblemData.MisclassificationMatrix[i - 1, i];
|
---|
[4366] | 136 | else
|
---|
[4391] | 137 | //true negative, consider only upper class
|
---|
[5319] | 138 | classificationScore += ProblemData.MisclassificationMatrix[i, i];
|
---|
[4366] | 139 | }
|
---|
| 140 | }
|
---|
[5341] | 141 |
|
---|
| 142 | //new best classification score found
|
---|
[4391] | 143 | if (classificationScore < bestClassificationScore) {
|
---|
| 144 | bestClassificationScore = classificationScore;
|
---|
[5332] | 145 | lowestBestThreshold = actualThreshold;
|
---|
| 146 | highestBestThreshold = actualThreshold;
|
---|
[5341] | 147 | seriesOfEqualClassificationScores = true;
|
---|
| 148 | }
|
---|
| 149 | //equal classification scores => if seriesOfEqualClassifcationScores == true update highest threshold
|
---|
| 150 | else if (Math.Abs(classificationScore - bestClassificationScore) < double.Epsilon && seriesOfEqualClassificationScores)
|
---|
[5332] | 151 | highestBestThreshold = actualThreshold;
|
---|
[5341] | 152 | //worse classificatoin score found reset seriesOfEqualClassifcationScores
|
---|
| 153 | else seriesOfEqualClassificationScores = false;
|
---|
[5332] | 154 |
|
---|
[4366] | 155 | actualThreshold += thresholdIncrement;
|
---|
| 156 | }
|
---|
[5341] | 157 | //scale lowest thresholds and highest found optimal threshold according to the misclassification matrix
|
---|
[5332] | 158 | double falseNegativePenalty = ProblemData.MisclassificationMatrix[i, i - 1];
|
---|
| 159 | double falsePositivePenalty = ProblemData.MisclassificationMatrix[i - 1, i];
|
---|
| 160 | thresholds[i] = (lowestBestThreshold * falsePositivePenalty + highestBestThreshold * falseNegativePenalty) / (falseNegativePenalty + falsePositivePenalty);
|
---|
[4366] | 161 | }
|
---|
| 162 | this.optimalThresholds = new List<double>(thresholds);
|
---|
| 163 | this.actualThresholds = optimalThresholds;
|
---|
| 164 | }
|
---|
| 165 |
|
---|
[4469] | 166 | public IEnumerable<double> GetEstimatedClassValues(IEnumerable<int> rows) {
|
---|
| 167 | double[] classValues = ProblemData.SortedClassValues.ToArray();
|
---|
[5321] | 168 | if (estimatedValues == null)
|
---|
| 169 | RecalculateEstimatedValues();
|
---|
[4469] | 170 | foreach (int row in rows) {
|
---|
| 171 | double value = estimatedValues[row];
|
---|
| 172 | int classIndex = 0;
|
---|
| 173 | while (value > actualThresholds[classIndex + 1])
|
---|
| 174 | classIndex++;
|
---|
| 175 | yield return classValues[classIndex];
|
---|
[4366] | 176 | }
|
---|
| 177 | }
|
---|
| 178 | #endregion
|
---|
| 179 |
|
---|
| 180 | public event EventHandler ThresholdsChanged;
|
---|
| 181 | private void OnThresholdsChanged() {
|
---|
| 182 | var handler = ThresholdsChanged;
|
---|
| 183 | if (handler != null)
|
---|
| 184 | ThresholdsChanged(this, EventArgs.Empty);
|
---|
| 185 | }
|
---|
| 186 | }
|
---|
| 187 | }
|
---|