| 1 | #region License Information
|
|---|
| 2 | /* HeuristicLab
|
|---|
| 3 | * Copyright (C) 2002-2011 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 |
|
|---|
| 22 | using System.Collections.Generic;
|
|---|
| 23 | using System.Linq;
|
|---|
| 24 | using HeuristicLab.Common;
|
|---|
| 25 | using HeuristicLab.Core;
|
|---|
| 26 | using HeuristicLab.Data;
|
|---|
| 27 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
|---|
| 28 | using HeuristicLab.Operators;
|
|---|
| 29 | using HeuristicLab.Optimization;
|
|---|
| 30 | using HeuristicLab.Parameters;
|
|---|
| 31 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
|---|
| 32 | using HeuristicLab.Problems.DataAnalysis.Evaluators;
|
|---|
| 33 | using HeuristicLab.Problems.DataAnalysis.Symbolic;
|
|---|
| 34 |
|
|---|
| 35 | namespace HeuristicLab.Problems.DataAnalysis.Regression.Symbolic.Analyzers {
|
|---|
| 36 | /// <summary>
|
|---|
| 37 | /// An operator that analyzes the training best scaled symbolic regression solution.
|
|---|
| 38 | /// </summary>
|
|---|
| 39 | [Item("TrainingBestScaledSymbolicRegressionSolutionAnalyzer", "An operator that analyzes the training best scaled symbolic regression solution.")]
|
|---|
| 40 | [StorableClass]
|
|---|
| 41 | public sealed class TrainingBestScaledSymbolicRegressionSolutionAnalyzer : SingleSuccessorOperator, ISymbolicRegressionAnalyzer {
|
|---|
| 42 | private const string ApplyLinearScalingParameterName = "ApplyLinearScaling";
|
|---|
| 43 | private const string SymbolicExpressionTreeParameterName = "SymbolicExpressionTree";
|
|---|
| 44 | private const string QualityParameterName = "Quality";
|
|---|
| 45 | private const string MaximizationParameterName = "Maximization";
|
|---|
| 46 | private const string CalculateSolutionComplexityParameterName = "CalculateSolutionComplexity";
|
|---|
| 47 | private const string CalculateSolutionAccuracyParameterName = "CalculateSolutionAccuracy";
|
|---|
| 48 | private const string SymbolicExpressionTreeInterpreterParameterName = "SymbolicExpressionTreeInterpreter";
|
|---|
| 49 | private const string ProblemDataParameterName = "DataAnalysisProblemData";
|
|---|
| 50 | private const string UpperEstimationLimitParameterName = "UpperEstimationLimit";
|
|---|
| 51 | private const string LowerEstimationLimitParameterName = "LowerEstimationLimit";
|
|---|
| 52 | private const string BestSolutionParameterName = "Best training solution";
|
|---|
| 53 | private const string BestSolutionQualityParameterName = "Best training solution quality";
|
|---|
| 54 | private const string BestSolutionLengthParameterName = "Best training solution length";
|
|---|
| 55 | private const string BestSolutionHeightParameterName = "Best training solution height";
|
|---|
| 56 | private const string BestSolutionVariablesParameterName = "Best training solution variables";
|
|---|
| 57 | private const string BestSolutionTrainingRSquaredParameterName = "Best training solution R² (training)";
|
|---|
| 58 | private const string BestSolutionTestRSquaredParameterName = "Best training solution R² (test)";
|
|---|
| 59 | private const string BestSolutionTrainingMseParameterName = "Best training solution mean squared error (training)";
|
|---|
| 60 | private const string BestSolutionTestMseParameterName = "Best training solution mean squared error (test)";
|
|---|
| 61 | private const string BestSolutionTrainingRelativeErrorParameterName = "Best training solution relative error (training)";
|
|---|
| 62 | private const string BestSolutionTestRelativeErrorParameterName = "Best training solution relative error (test)";
|
|---|
| 63 | private const string ResultsParameterName = "Results";
|
|---|
| 64 |
|
|---|
| 65 | #region parameter properties
|
|---|
| 66 | public ScopeTreeLookupParameter<SymbolicExpressionTree> SymbolicExpressionTreeParameter {
|
|---|
| 67 | get { return (ScopeTreeLookupParameter<SymbolicExpressionTree>)Parameters[SymbolicExpressionTreeParameterName]; }
|
|---|
| 68 | }
|
|---|
| 69 | public ScopeTreeLookupParameter<DoubleValue> QualityParameter {
|
|---|
| 70 | get { return (ScopeTreeLookupParameter<DoubleValue>)Parameters[QualityParameterName]; }
|
|---|
| 71 | }
|
|---|
| 72 | public ILookupParameter<BoolValue> MaximizationParameter {
|
|---|
| 73 | get { return (ILookupParameter<BoolValue>)Parameters[MaximizationParameterName]; }
|
|---|
| 74 | }
|
|---|
| 75 | public IValueParameter<BoolValue> CalculateSolutionComplexityParameter {
|
|---|
| 76 | get { return (IValueParameter<BoolValue>)Parameters[CalculateSolutionComplexityParameterName]; }
|
|---|
| 77 | }
|
|---|
| 78 | public IValueParameter<BoolValue> CalculateSolutionAccuracyParameter {
|
|---|
| 79 | get { return (IValueParameter<BoolValue>)Parameters[CalculateSolutionAccuracyParameterName]; }
|
|---|
| 80 | }
|
|---|
| 81 | public IValueLookupParameter<ISymbolicExpressionTreeInterpreter> SymbolicExpressionTreeInterpreterParameter {
|
|---|
| 82 | get { return (IValueLookupParameter<ISymbolicExpressionTreeInterpreter>)Parameters[SymbolicExpressionTreeInterpreterParameterName]; }
|
|---|
| 83 | }
|
|---|
| 84 | public IValueLookupParameter<DataAnalysisProblemData> ProblemDataParameter {
|
|---|
| 85 | get { return (IValueLookupParameter<DataAnalysisProblemData>)Parameters[ProblemDataParameterName]; }
|
|---|
| 86 | }
|
|---|
| 87 | public IValueLookupParameter<DoubleValue> UpperEstimationLimitParameter {
|
|---|
| 88 | get { return (IValueLookupParameter<DoubleValue>)Parameters[UpperEstimationLimitParameterName]; }
|
|---|
| 89 | }
|
|---|
| 90 | public IValueLookupParameter<DoubleValue> LowerEstimationLimitParameter {
|
|---|
| 91 | get { return (IValueLookupParameter<DoubleValue>)Parameters[LowerEstimationLimitParameterName]; }
|
|---|
| 92 | }
|
|---|
| 93 |
|
|---|
| 94 | public ILookupParameter<SymbolicRegressionSolution> BestSolutionParameter {
|
|---|
| 95 | get { return (ILookupParameter<SymbolicRegressionSolution>)Parameters[BestSolutionParameterName]; }
|
|---|
| 96 | }
|
|---|
| 97 | public ILookupParameter<DoubleValue> BestSolutionQualityParameter {
|
|---|
| 98 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionQualityParameterName]; }
|
|---|
| 99 | }
|
|---|
| 100 | public ILookupParameter<IntValue> BestSolutionLengthParameter {
|
|---|
| 101 | get { return (ILookupParameter<IntValue>)Parameters[BestSolutionLengthParameterName]; }
|
|---|
| 102 | }
|
|---|
| 103 | public ILookupParameter<IntValue> BestSolutionHeightParameter {
|
|---|
| 104 | get { return (ILookupParameter<IntValue>)Parameters[BestSolutionHeightParameterName]; }
|
|---|
| 105 | }
|
|---|
| 106 | public ILookupParameter<IntValue> BestSolutionVariablesParameter {
|
|---|
| 107 | get { return (ILookupParameter<IntValue>)Parameters[BestSolutionVariablesParameterName]; }
|
|---|
| 108 | }
|
|---|
| 109 | public ILookupParameter<DoubleValue> BestSolutionTrainingRSquaredParameter {
|
|---|
| 110 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTrainingRSquaredParameterName]; }
|
|---|
| 111 | }
|
|---|
| 112 | public ILookupParameter<DoubleValue> BestSolutionTestRSquaredParameter {
|
|---|
| 113 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTestRSquaredParameterName]; }
|
|---|
| 114 | }
|
|---|
| 115 | public ILookupParameter<DoubleValue> BestSolutionTrainingMseParameter {
|
|---|
| 116 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTrainingMseParameterName]; }
|
|---|
| 117 | }
|
|---|
| 118 | public ILookupParameter<DoubleValue> BestSolutionTestMseParameter {
|
|---|
| 119 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTestMseParameterName]; }
|
|---|
| 120 | }
|
|---|
| 121 | public ILookupParameter<DoubleValue> BestSolutionTrainingRelativeErrorParameter {
|
|---|
| 122 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTrainingRelativeErrorParameterName]; }
|
|---|
| 123 | }
|
|---|
| 124 | public ILookupParameter<DoubleValue> BestSolutionTestRelativeErrorParameter {
|
|---|
| 125 | get { return (ILookupParameter<DoubleValue>)Parameters[BestSolutionTestRelativeErrorParameterName]; }
|
|---|
| 126 | }
|
|---|
| 127 | public ILookupParameter<ResultCollection> ResultsParameter {
|
|---|
| 128 | get { return (ILookupParameter<ResultCollection>)Parameters[ResultsParameterName]; }
|
|---|
| 129 | }
|
|---|
| 130 | public IValueLookupParameter<BoolValue> ApplyLinearScalingParameter {
|
|---|
| 131 | get { return (IValueLookupParameter<BoolValue>)Parameters[ApplyLinearScalingParameterName]; }
|
|---|
| 132 | }
|
|---|
| 133 | #endregion
|
|---|
| 134 | #region properties
|
|---|
| 135 | public ItemArray<SymbolicExpressionTree> SymbolicExpressionTree {
|
|---|
| 136 | get { return SymbolicExpressionTreeParameter.ActualValue; }
|
|---|
| 137 | }
|
|---|
| 138 | public ItemArray<DoubleValue> Quality {
|
|---|
| 139 | get { return QualityParameter.ActualValue; }
|
|---|
| 140 | }
|
|---|
| 141 | public BoolValue Maximization {
|
|---|
| 142 | get { return MaximizationParameter.ActualValue; }
|
|---|
| 143 | }
|
|---|
| 144 | public BoolValue CalculateSolutionComplexity {
|
|---|
| 145 | get { return CalculateSolutionComplexityParameter.Value; }
|
|---|
| 146 | set { CalculateSolutionComplexityParameter.Value = value; }
|
|---|
| 147 | }
|
|---|
| 148 | public BoolValue CalculateSolutionAccuracy {
|
|---|
| 149 | get { return CalculateSolutionAccuracyParameter.Value; }
|
|---|
| 150 | set { CalculateSolutionAccuracyParameter.Value = value; }
|
|---|
| 151 | }
|
|---|
| 152 | public ISymbolicExpressionTreeInterpreter SymbolicExpressionTreeInterpreter {
|
|---|
| 153 | get { return SymbolicExpressionTreeInterpreterParameter.ActualValue; }
|
|---|
| 154 | }
|
|---|
| 155 | public DataAnalysisProblemData ProblemData {
|
|---|
| 156 | get { return ProblemDataParameter.ActualValue; }
|
|---|
| 157 | }
|
|---|
| 158 | public DoubleValue UpperEstimationLimit {
|
|---|
| 159 | get { return UpperEstimationLimitParameter.ActualValue; }
|
|---|
| 160 | }
|
|---|
| 161 | public DoubleValue LowerEstimationLimit {
|
|---|
| 162 | get { return LowerEstimationLimitParameter.ActualValue; }
|
|---|
| 163 | }
|
|---|
| 164 | public ResultCollection Results {
|
|---|
| 165 | get { return ResultsParameter.ActualValue; }
|
|---|
| 166 | }
|
|---|
| 167 | public SymbolicRegressionSolution BestSolution {
|
|---|
| 168 | get { return BestSolutionParameter.ActualValue; }
|
|---|
| 169 | set { BestSolutionParameter.ActualValue = value; }
|
|---|
| 170 | }
|
|---|
| 171 | public DoubleValue BestSolutionQuality {
|
|---|
| 172 | get { return BestSolutionQualityParameter.ActualValue; }
|
|---|
| 173 | set { BestSolutionQualityParameter.ActualValue = value; }
|
|---|
| 174 | }
|
|---|
| 175 | public IntValue BestSolutionLength {
|
|---|
| 176 | get { return BestSolutionLengthParameter.ActualValue; }
|
|---|
| 177 | set { BestSolutionLengthParameter.ActualValue = value; }
|
|---|
| 178 | }
|
|---|
| 179 | public IntValue BestSolutionHeight {
|
|---|
| 180 | get { return BestSolutionHeightParameter.ActualValue; }
|
|---|
| 181 | set { BestSolutionHeightParameter.ActualValue = value; }
|
|---|
| 182 | }
|
|---|
| 183 | public IntValue BestSolutionVariables {
|
|---|
| 184 | get { return BestSolutionVariablesParameter.ActualValue; }
|
|---|
| 185 | set { BestSolutionVariablesParameter.ActualValue = value; }
|
|---|
| 186 | }
|
|---|
| 187 | public DoubleValue BestSolutionTrainingRSquared {
|
|---|
| 188 | get { return BestSolutionTrainingRSquaredParameter.ActualValue; }
|
|---|
| 189 | set { BestSolutionTrainingRSquaredParameter.ActualValue = value; }
|
|---|
| 190 | }
|
|---|
| 191 | public DoubleValue BestSolutionTestRSquared {
|
|---|
| 192 | get { return BestSolutionTestRSquaredParameter.ActualValue; }
|
|---|
| 193 | set { BestSolutionTestRSquaredParameter.ActualValue = value; }
|
|---|
| 194 | }
|
|---|
| 195 | public DoubleValue BestSolutionTrainingMse {
|
|---|
| 196 | get { return BestSolutionTrainingMseParameter.ActualValue; }
|
|---|
| 197 | set { BestSolutionTrainingMseParameter.ActualValue = value; }
|
|---|
| 198 | }
|
|---|
| 199 | public DoubleValue BestSolutionTestMse {
|
|---|
| 200 | get { return BestSolutionTestMseParameter.ActualValue; }
|
|---|
| 201 | set { BestSolutionTestMseParameter.ActualValue = value; }
|
|---|
| 202 | }
|
|---|
| 203 | public DoubleValue BestSolutionTrainingRelativeError {
|
|---|
| 204 | get { return BestSolutionTrainingRelativeErrorParameter.ActualValue; }
|
|---|
| 205 | set { BestSolutionTrainingRelativeErrorParameter.ActualValue = value; }
|
|---|
| 206 | }
|
|---|
| 207 | public DoubleValue BestSolutionTestRelativeError {
|
|---|
| 208 | get { return BestSolutionTestRelativeErrorParameter.ActualValue; }
|
|---|
| 209 | set { BestSolutionTestRelativeErrorParameter.ActualValue = value; }
|
|---|
| 210 | }
|
|---|
| 211 | public BoolValue ApplyLinearScaling {
|
|---|
| 212 | get { return ApplyLinearScalingParameter.ActualValue; }
|
|---|
| 213 | set { ApplyLinearScalingParameter.ActualValue = value; }
|
|---|
| 214 | }
|
|---|
| 215 | #endregion
|
|---|
| 216 |
|
|---|
| 217 | [StorableConstructor]
|
|---|
| 218 | private TrainingBestScaledSymbolicRegressionSolutionAnalyzer(bool deserializing) : base(deserializing) { }
|
|---|
| 219 | private TrainingBestScaledSymbolicRegressionSolutionAnalyzer(TrainingBestScaledSymbolicRegressionSolutionAnalyzer original, Cloner cloner) : base(original, cloner) { }
|
|---|
| 220 | public TrainingBestScaledSymbolicRegressionSolutionAnalyzer()
|
|---|
| 221 | : base() {
|
|---|
| 222 | Parameters.Add(new ValueLookupParameter<BoolValue>(ApplyLinearScalingParameterName, "The switch determines if the best solution should be linearly scaled on the whole training set.", new BoolValue(true)));
|
|---|
| 223 | Parameters.Add(new LookupParameter<BoolValue>(MaximizationParameterName, "The direction of optimization."));
|
|---|
| 224 | Parameters.Add(new ScopeTreeLookupParameter<SymbolicExpressionTree>(SymbolicExpressionTreeParameterName, "The symbolic expression trees to analyze."));
|
|---|
| 225 | Parameters.Add(new ScopeTreeLookupParameter<DoubleValue>(QualityParameterName, "The qualities of the symbolic expression trees to analyze."));
|
|---|
| 226 | Parameters.Add(new ValueParameter<BoolValue>(CalculateSolutionComplexityParameterName, "Determines if the length and height of the training best solution should be calculated.", new BoolValue(true)));
|
|---|
| 227 | Parameters.Add(new ValueParameter<BoolValue>(CalculateSolutionAccuracyParameterName, "Determines if the accuracy of the training best solution on the training and test set should be calculated.", new BoolValue(true)));
|
|---|
| 228 | Parameters.Add(new ValueLookupParameter<ISymbolicExpressionTreeInterpreter>(SymbolicExpressionTreeInterpreterParameterName, "The interpreter that should be used for the analysis of symbolic expression trees."));
|
|---|
| 229 | Parameters.Add(new ValueLookupParameter<DataAnalysisProblemData>(ProblemDataParameterName, "The problem data for which the symbolic expression tree is a solution."));
|
|---|
| 230 | Parameters.Add(new ValueLookupParameter<DoubleValue>(UpperEstimationLimitParameterName, "The upper estimation limit that was set for the evaluation of the symbolic expression trees."));
|
|---|
| 231 | Parameters.Add(new ValueLookupParameter<DoubleValue>(LowerEstimationLimitParameterName, "The lower estimation limit that was set for the evaluation of the symbolic expression trees."));
|
|---|
| 232 | Parameters.Add(new LookupParameter<SymbolicRegressionSolution>(BestSolutionParameterName, "The best symbolic regression solution."));
|
|---|
| 233 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionQualityParameterName, "The quality of the best symbolic regression solution."));
|
|---|
| 234 | Parameters.Add(new LookupParameter<IntValue>(BestSolutionLengthParameterName, "The length of the best symbolic regression solution."));
|
|---|
| 235 | Parameters.Add(new LookupParameter<IntValue>(BestSolutionHeightParameterName, "The height of the best symbolic regression solution."));
|
|---|
| 236 | Parameters.Add(new LookupParameter<IntValue>(BestSolutionVariablesParameterName, "The number of variables used by the best symbolic regression solution."));
|
|---|
| 237 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTrainingRSquaredParameterName, "The R² value on the training set of the best symbolic regression solution."));
|
|---|
| 238 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTestRSquaredParameterName, "The R² value on the test set of the best symbolic regression solution."));
|
|---|
| 239 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTrainingMseParameterName, "The mean squared error on the training set of the best symbolic regression solution."));
|
|---|
| 240 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTestMseParameterName, "The mean squared error value on the test set of the best symbolic regression solution."));
|
|---|
| 241 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTrainingRelativeErrorParameterName, "The relative error on the training set of the best symbolic regression solution."));
|
|---|
| 242 | Parameters.Add(new LookupParameter<DoubleValue>(BestSolutionTestRelativeErrorParameterName, "The relative error value on the test set of the best symbolic regression solution."));
|
|---|
| 243 | Parameters.Add(new LookupParameter<ResultCollection>(ResultsParameterName, "The result collection where the best symbolic regression solution should be stored."));
|
|---|
| 244 | }
|
|---|
| 245 |
|
|---|
| 246 | public override IDeepCloneable Clone(Cloner cloner) {
|
|---|
| 247 | return new TrainingBestScaledSymbolicRegressionSolutionAnalyzer(this, cloner);
|
|---|
| 248 | }
|
|---|
| 249 |
|
|---|
| 250 | [StorableHook(HookType.AfterDeserialization)]
|
|---|
| 251 | private void AfterDeserialization() {
|
|---|
| 252 | if (!Parameters.ContainsKey(ApplyLinearScalingParameterName)) {
|
|---|
| 253 | Parameters.Add(new ValueLookupParameter<BoolValue>(ApplyLinearScalingParameterName, "The switch determines if the best solution should be linearly scaled on the whole training set.", new BoolValue(true)));
|
|---|
| 254 | }
|
|---|
| 255 | }
|
|---|
| 256 |
|
|---|
| 257 | public override IOperation Apply() {
|
|---|
| 258 | #region find best tree
|
|---|
| 259 | double bestQuality = Maximization.Value ? double.NegativeInfinity : double.PositiveInfinity;
|
|---|
| 260 | SymbolicExpressionTree bestTree = null;
|
|---|
| 261 | SymbolicExpressionTree[] tree = SymbolicExpressionTree.ToArray();
|
|---|
| 262 | double[] quality = Quality.Select(x => x.Value).ToArray();
|
|---|
| 263 | for (int i = 0; i < tree.Length; i++) {
|
|---|
| 264 | if ((Maximization.Value && quality[i] > bestQuality) ||
|
|---|
| 265 | (!Maximization.Value && quality[i] < bestQuality)) {
|
|---|
| 266 | bestQuality = quality[i];
|
|---|
| 267 | bestTree = tree[i];
|
|---|
| 268 | }
|
|---|
| 269 | }
|
|---|
| 270 | #endregion
|
|---|
| 271 |
|
|---|
| 272 | #region update best solution
|
|---|
| 273 | // if the best tree is better than the current best solution => update
|
|---|
| 274 | bool newBest =
|
|---|
| 275 | BestSolutionQuality == null ||
|
|---|
| 276 | (Maximization.Value && bestQuality > BestSolutionQuality.Value) ||
|
|---|
| 277 | (!Maximization.Value && bestQuality < BestSolutionQuality.Value);
|
|---|
| 278 | if (newBest) {
|
|---|
| 279 | double upperEstimationLimit = UpperEstimationLimit != null ? UpperEstimationLimit.Value : double.PositiveInfinity;
|
|---|
| 280 | double lowerEstimationLimit = LowerEstimationLimit != null ? LowerEstimationLimit.Value : double.NegativeInfinity;
|
|---|
| 281 | string targetVariable = ProblemData.TargetVariable.Value;
|
|---|
| 282 |
|
|---|
| 283 | if (ApplyLinearScaling.Value) {
|
|---|
| 284 | // calculate scaling parameters and only for the best tree using the full training set
|
|---|
| 285 | double alpha, beta;
|
|---|
| 286 | SymbolicRegressionScaledMeanSquaredErrorEvaluator.Calculate(SymbolicExpressionTreeInterpreter, bestTree,
|
|---|
| 287 | lowerEstimationLimit, upperEstimationLimit,
|
|---|
| 288 | ProblemData.Dataset, targetVariable,
|
|---|
| 289 | ProblemData.TrainingIndizes, out beta, out alpha);
|
|---|
| 290 |
|
|---|
| 291 | // scale tree for solution
|
|---|
| 292 | bestTree = SymbolicRegressionSolutionLinearScaler.Scale(bestTree, alpha, beta);
|
|---|
| 293 | }
|
|---|
| 294 | var model = new SymbolicRegressionModel((ISymbolicExpressionTreeInterpreter)SymbolicExpressionTreeInterpreter.Clone(),
|
|---|
| 295 | bestTree);
|
|---|
| 296 | var solution = new SymbolicRegressionSolution((DataAnalysisProblemData)ProblemData.Clone(), model, lowerEstimationLimit, upperEstimationLimit);
|
|---|
| 297 | solution.Name = BestSolutionParameterName;
|
|---|
| 298 | solution.Description = "Best solution on training partition found over the whole run.";
|
|---|
| 299 |
|
|---|
| 300 | BestSolution = solution;
|
|---|
| 301 | BestSolutionQuality = new DoubleValue(bestQuality);
|
|---|
| 302 |
|
|---|
| 303 | if (CalculateSolutionComplexity.Value) {
|
|---|
| 304 | BestSolutionLength = new IntValue(solution.Model.SymbolicExpressionTree.Size);
|
|---|
| 305 | BestSolutionHeight = new IntValue(solution.Model.SymbolicExpressionTree.Height);
|
|---|
| 306 | BestSolutionVariables = new IntValue(solution.Model.InputVariables.Count());
|
|---|
| 307 | if (!Results.ContainsKey(BestSolutionLengthParameterName)) {
|
|---|
| 308 | Results.Add(new Result(BestSolutionLengthParameterName, "Length of the best solution on the training set.", BestSolutionLength));
|
|---|
| 309 | Results.Add(new Result(BestSolutionHeightParameterName, "Height of the best solution on the training set.", BestSolutionHeight));
|
|---|
| 310 | Results.Add(new Result(BestSolutionVariablesParameterName, "Number of variables used by the best solution on the training set.", BestSolutionVariables));
|
|---|
| 311 | } else {
|
|---|
| 312 | Results[BestSolutionLengthParameterName].Value = BestSolutionLength;
|
|---|
| 313 | Results[BestSolutionHeightParameterName].Value = BestSolutionHeight;
|
|---|
| 314 | Results[BestSolutionVariablesParameterName].Value = BestSolutionVariables;
|
|---|
| 315 | }
|
|---|
| 316 | }
|
|---|
| 317 |
|
|---|
| 318 | if (CalculateSolutionAccuracy.Value) {
|
|---|
| 319 | #region update R2,MSE, Rel Error
|
|---|
| 320 | IEnumerable<double> trainingValues = ProblemData.Dataset.GetEnumeratedVariableValues(ProblemData.TargetVariable.Value, ProblemData.TrainingIndizes);
|
|---|
| 321 | IEnumerable<double> testValues = ProblemData.Dataset.GetEnumeratedVariableValues(ProblemData.TargetVariable.Value, ProblemData.TestIndizes);
|
|---|
| 322 | OnlineMeanSquaredErrorEvaluator mseEvaluator = new OnlineMeanSquaredErrorEvaluator();
|
|---|
| 323 | OnlineMeanAbsolutePercentageErrorEvaluator relErrorEvaluator = new OnlineMeanAbsolutePercentageErrorEvaluator();
|
|---|
| 324 | OnlinePearsonsRSquaredEvaluator r2Evaluator = new OnlinePearsonsRSquaredEvaluator();
|
|---|
| 325 |
|
|---|
| 326 | #region training
|
|---|
| 327 | var originalEnumerator = trainingValues.GetEnumerator();
|
|---|
| 328 | var estimatedEnumerator = solution.EstimatedTrainingValues.GetEnumerator();
|
|---|
| 329 | while (originalEnumerator.MoveNext() & estimatedEnumerator.MoveNext()) {
|
|---|
| 330 | mseEvaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 331 | r2Evaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 332 | relErrorEvaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 333 | }
|
|---|
| 334 | double trainingR2 = r2Evaluator.RSquared;
|
|---|
| 335 | double trainingMse = mseEvaluator.MeanSquaredError;
|
|---|
| 336 | double trainingRelError = relErrorEvaluator.MeanAbsolutePercentageError;
|
|---|
| 337 | #endregion
|
|---|
| 338 |
|
|---|
| 339 | mseEvaluator.Reset();
|
|---|
| 340 | relErrorEvaluator.Reset();
|
|---|
| 341 | r2Evaluator.Reset();
|
|---|
| 342 |
|
|---|
| 343 | #region test
|
|---|
| 344 | originalEnumerator = testValues.GetEnumerator();
|
|---|
| 345 | estimatedEnumerator = solution.EstimatedTestValues.GetEnumerator();
|
|---|
| 346 | while (originalEnumerator.MoveNext() & estimatedEnumerator.MoveNext()) {
|
|---|
| 347 | mseEvaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 348 | r2Evaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 349 | relErrorEvaluator.Add(originalEnumerator.Current, estimatedEnumerator.Current);
|
|---|
| 350 | }
|
|---|
| 351 | double testR2 = r2Evaluator.RSquared;
|
|---|
| 352 | double testMse = mseEvaluator.MeanSquaredError;
|
|---|
| 353 | double testRelError = relErrorEvaluator.MeanAbsolutePercentageError;
|
|---|
| 354 | #endregion
|
|---|
| 355 | BestSolutionTrainingRSquared = new DoubleValue(trainingR2);
|
|---|
| 356 | BestSolutionTestRSquared = new DoubleValue(testR2);
|
|---|
| 357 | BestSolutionTrainingMse = new DoubleValue(trainingMse);
|
|---|
| 358 | BestSolutionTestMse = new DoubleValue(testMse);
|
|---|
| 359 | BestSolutionTrainingRelativeError = new DoubleValue(trainingRelError);
|
|---|
| 360 | BestSolutionTestRelativeError = new DoubleValue(testRelError);
|
|---|
| 361 |
|
|---|
| 362 | if (!Results.ContainsKey(BestSolutionTrainingRSquaredParameterName)) {
|
|---|
| 363 | Results.Add(new Result(BestSolutionTrainingRSquaredParameterName, BestSolutionTrainingRSquared));
|
|---|
| 364 | Results.Add(new Result(BestSolutionTestRSquaredParameterName, BestSolutionTestRSquared));
|
|---|
| 365 | Results.Add(new Result(BestSolutionTrainingMseParameterName, BestSolutionTrainingMse));
|
|---|
| 366 | Results.Add(new Result(BestSolutionTestMseParameterName, BestSolutionTestMse));
|
|---|
| 367 | Results.Add(new Result(BestSolutionTrainingRelativeErrorParameterName, BestSolutionTrainingRelativeError));
|
|---|
| 368 | Results.Add(new Result(BestSolutionTestRelativeErrorParameterName, BestSolutionTestRelativeError));
|
|---|
| 369 | } else {
|
|---|
| 370 | Results[BestSolutionTrainingRSquaredParameterName].Value = BestSolutionTrainingRSquared;
|
|---|
| 371 | Results[BestSolutionTestRSquaredParameterName].Value = BestSolutionTestRSquared;
|
|---|
| 372 | Results[BestSolutionTrainingMseParameterName].Value = BestSolutionTrainingMse;
|
|---|
| 373 | Results[BestSolutionTestMseParameterName].Value = BestSolutionTestMse;
|
|---|
| 374 | Results[BestSolutionTrainingRelativeErrorParameterName].Value = BestSolutionTrainingRelativeError;
|
|---|
| 375 | Results[BestSolutionTestRelativeErrorParameterName].Value = BestSolutionTestRelativeError;
|
|---|
| 376 | }
|
|---|
| 377 | #endregion
|
|---|
| 378 | }
|
|---|
| 379 |
|
|---|
| 380 | if (!Results.ContainsKey(BestSolutionQualityParameterName)) {
|
|---|
| 381 | Results.Add(new Result(BestSolutionQualityParameterName, BestSolutionQuality));
|
|---|
| 382 | Results.Add(new Result(BestSolutionParameterName, BestSolution));
|
|---|
| 383 | } else {
|
|---|
| 384 | Results[BestSolutionQualityParameterName].Value = BestSolutionQuality;
|
|---|
| 385 | Results[BestSolutionParameterName].Value = BestSolution;
|
|---|
| 386 | }
|
|---|
| 387 | }
|
|---|
| 388 | #endregion
|
|---|
| 389 | return base.Apply();
|
|---|
| 390 | }
|
|---|
| 391 | }
|
|---|
| 392 | }
|
|---|
