| 1 | #region License Information
|
|---|
| 2 | /* HeuristicLab
|
|---|
| 3 | * Copyright (C) 2002-2010 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.Linq;
|
|---|
| 23 | using HeuristicLab.Common;
|
|---|
| 24 | using HeuristicLab.Core;
|
|---|
| 25 | using HeuristicLab.Data;
|
|---|
| 26 | using HeuristicLab.Operators;
|
|---|
| 27 | using HeuristicLab.Optimization;
|
|---|
| 28 | using HeuristicLab.Parameters;
|
|---|
| 29 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
|---|
| 30 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
|---|
| 31 | using HeuristicLab.Problems.DataAnalysis.Evaluators;
|
|---|
| 32 | using HeuristicLab.Problems.DataAnalysis.Symbolic;
|
|---|
| 33 | using System.Collections.Generic;
|
|---|
| 34 | using HeuristicLab.Analysis;
|
|---|
| 35 |
|
|---|
| 36 | using HeuristicLab.Problems.DataAnalysis.Symbolic.Symbols;
|
|---|
| 37 |
|
|---|
| 38 | namespace HeuristicLab.Problems.DataAnalysis.Regression.Symbolic {
|
|---|
| 39 | /// <summary>
|
|---|
| 40 | /// An operator for visualizing the best symbolic regression solution based on the validation set.
|
|---|
| 41 | /// </summary>
|
|---|
| 42 | [Item("BestSymbolicExpressionTreeVisualizer", "An operator for visualizing the best symbolic regression solution based on the validation set.")]
|
|---|
| 43 | [StorableClass]
|
|---|
| 44 | public sealed class BestValidationSymbolicRegressionSolutionVisualizer : SingleSuccessorOperator, IAnalyzer {
|
|---|
| 45 | private const string SymbolicExpressionTreeInterpreterParameterName = "SymbolicExpressionTreeInterpreter";
|
|---|
| 46 | private const string UpperEstimationLimitParameterName = "UpperEstimationLimit";
|
|---|
| 47 | private const string LowerEstimationLimitParameterName = "LowerEstimationLimit";
|
|---|
| 48 | private const string AlphaParameterName = "Alpha";
|
|---|
| 49 | private const string BetaParameterName = "Beta";
|
|---|
| 50 | private const string SymbolicRegressionModelParameterName = "SymbolicRegressionModel";
|
|---|
| 51 | private const string DataAnalysisProblemDataParameterName = "DataAnalysisProblemData";
|
|---|
| 52 | private const string BestValidationSolutionParameterName = "BestValidationSolution";
|
|---|
| 53 | private const string ValidationSamplesStartParameterName = "ValidationSamplesStart";
|
|---|
| 54 | private const string ValidationSamplesEndParameterName = "ValidationSamplesEnd";
|
|---|
| 55 | private const string QualityParameterName = "Quality";
|
|---|
| 56 | private const string ResultsParameterName = "Results";
|
|---|
| 57 | private const string VariableFrequenciesParameterName = "VariableFrequencies";
|
|---|
| 58 |
|
|---|
| 59 | #region parameter properties
|
|---|
| 60 | public ILookupParameter<ISymbolicExpressionTreeInterpreter> SymbolicExpressionTreeInterpreterParameter {
|
|---|
| 61 | get { return (ILookupParameter<ISymbolicExpressionTreeInterpreter>)Parameters[SymbolicExpressionTreeInterpreterParameterName]; }
|
|---|
| 62 | }
|
|---|
| 63 | public IValueLookupParameter<DoubleValue> UpperEstimationLimitParameter {
|
|---|
| 64 | get { return (IValueLookupParameter<DoubleValue>)Parameters[UpperEstimationLimitParameterName]; }
|
|---|
| 65 | }
|
|---|
| 66 | public IValueLookupParameter<DoubleValue> LowerEstimationLimitParameter {
|
|---|
| 67 | get { return (IValueLookupParameter<DoubleValue>)Parameters[LowerEstimationLimitParameterName]; }
|
|---|
| 68 | }
|
|---|
| 69 | public IValueLookupParameter<IntValue> ValidationSamplesStartParameter {
|
|---|
| 70 | get { return (IValueLookupParameter<IntValue>)Parameters[ValidationSamplesStartParameterName]; }
|
|---|
| 71 | }
|
|---|
| 72 | public IValueLookupParameter<IntValue> ValidationSamplesEndParameter {
|
|---|
| 73 | get { return (IValueLookupParameter<IntValue>)Parameters[ValidationSamplesEndParameterName]; }
|
|---|
| 74 | }
|
|---|
| 75 |
|
|---|
| 76 | public ILookupParameter<ItemArray<SymbolicExpressionTree>> SymbolicExpressionTreeParameter {
|
|---|
| 77 | get { return (ILookupParameter<ItemArray<SymbolicExpressionTree>>)Parameters[SymbolicRegressionModelParameterName]; }
|
|---|
| 78 | }
|
|---|
| 79 | public ILookupParameter<ItemArray<DoubleValue>> AlphaParameter {
|
|---|
| 80 | get { return (ILookupParameter<ItemArray<DoubleValue>>)Parameters[AlphaParameterName]; }
|
|---|
| 81 | }
|
|---|
| 82 | public ILookupParameter<ItemArray<DoubleValue>> BetaParameter {
|
|---|
| 83 | get { return (ILookupParameter<ItemArray<DoubleValue>>)Parameters[BetaParameterName]; }
|
|---|
| 84 | }
|
|---|
| 85 | public ILookupParameter<DataAnalysisProblemData> DataAnalysisProblemDataParameter {
|
|---|
| 86 | get { return (ILookupParameter<DataAnalysisProblemData>)Parameters[DataAnalysisProblemDataParameterName]; }
|
|---|
| 87 | }
|
|---|
| 88 | public ILookupParameter<SymbolicRegressionSolution> BestValidationSolutionParameter {
|
|---|
| 89 | get { return (ILookupParameter<SymbolicRegressionSolution>)Parameters[BestValidationSolutionParameterName]; }
|
|---|
| 90 | }
|
|---|
| 91 |
|
|---|
| 92 | public ILookupParameter<ItemArray<DoubleValue>> QualityParameter {
|
|---|
| 93 | get { return (ILookupParameter<ItemArray<DoubleValue>>)Parameters[QualityParameterName]; }
|
|---|
| 94 | }
|
|---|
| 95 |
|
|---|
| 96 | public ILookupParameter<ResultCollection> ResultParameter {
|
|---|
| 97 | get { return (ILookupParameter<ResultCollection>)Parameters[ResultsParameterName]; }
|
|---|
| 98 | }
|
|---|
| 99 | public ILookupParameter<DataTable> VariableFrequenciesParameter {
|
|---|
| 100 | get { return (ILookupParameter<DataTable>)Parameters[VariableFrequenciesParameterName]; }
|
|---|
| 101 | }
|
|---|
| 102 |
|
|---|
| 103 | #endregion
|
|---|
| 104 |
|
|---|
| 105 | #region properties
|
|---|
| 106 | public ISymbolicExpressionTreeInterpreter SymbolicExpressionTreeInterpreter {
|
|---|
| 107 | get { return SymbolicExpressionTreeInterpreterParameter.ActualValue; }
|
|---|
| 108 | }
|
|---|
| 109 | public DoubleValue UpperEstimationLimit {
|
|---|
| 110 | get { return UpperEstimationLimitParameter.ActualValue; }
|
|---|
| 111 | }
|
|---|
| 112 | public DoubleValue LowerEstimationLimit {
|
|---|
| 113 | get { return LowerEstimationLimitParameter.ActualValue; }
|
|---|
| 114 | }
|
|---|
| 115 | public IntValue ValidationSamplesStart {
|
|---|
| 116 | get { return ValidationSamplesStartParameter.ActualValue; }
|
|---|
| 117 | }
|
|---|
| 118 | public IntValue ValidationSamplesEnd {
|
|---|
| 119 | get { return ValidationSamplesEndParameter.ActualValue; }
|
|---|
| 120 | }
|
|---|
| 121 | public DataTable VariableFrequencies {
|
|---|
| 122 | get { return VariableFrequenciesParameter.ActualValue; }
|
|---|
| 123 | set { VariableFrequenciesParameter.ActualValue = value; }
|
|---|
| 124 | }
|
|---|
| 125 | #endregion
|
|---|
| 126 |
|
|---|
| 127 | public BestValidationSymbolicRegressionSolutionVisualizer()
|
|---|
| 128 | : base() {
|
|---|
| 129 | Parameters.Add(new SubScopesLookupParameter<SymbolicExpressionTree>(SymbolicRegressionModelParameterName, "The symbolic regression solutions from which the best solution should be visualized."));
|
|---|
| 130 | Parameters.Add(new SubScopesLookupParameter<DoubleValue>(QualityParameterName, "The quality of the symbolic regression solutions."));
|
|---|
| 131 | Parameters.Add(new LookupParameter<DataAnalysisProblemData>(DataAnalysisProblemDataParameterName, "The symbolic regression problme data on which the best solution should be evaluated."));
|
|---|
| 132 | Parameters.Add(new LookupParameter<ISymbolicExpressionTreeInterpreter>(SymbolicExpressionTreeInterpreterParameterName, "The interpreter that should be used to calculate the output values of symbolic expression trees."));
|
|---|
| 133 | Parameters.Add(new SubScopesLookupParameter<DoubleValue>(AlphaParameterName, "Alpha parameter for linear scaling of the estimated values."));
|
|---|
| 134 | Parameters.Add(new SubScopesLookupParameter<DoubleValue>(BetaParameterName, "Beta parameter for linear scaling ot the estimated values."));
|
|---|
| 135 | Parameters.Add(new ValueLookupParameter<DoubleValue>(UpperEstimationLimitParameterName, "The upper limit that should be used as cut off value for the output values of symbolic expression trees."));
|
|---|
| 136 | Parameters.Add(new ValueLookupParameter<DoubleValue>(LowerEstimationLimitParameterName, "The lower limit that should be used as cut off value for the output values of symbolic expression trees."));
|
|---|
| 137 | Parameters.Add(new ValueLookupParameter<IntValue>(ValidationSamplesStartParameterName, "The start index of the validation partition (part of the training partition)."));
|
|---|
| 138 | Parameters.Add(new ValueLookupParameter<IntValue>(ValidationSamplesEndParameterName, "The end index of the validation partition (part of the training partition)."));
|
|---|
| 139 | Parameters.Add(new LookupParameter<SymbolicRegressionSolution>(BestValidationSolutionParameterName, "The best symbolic expression tree based on the validation data for the symbolic regression problem."));
|
|---|
| 140 | Parameters.Add(new LookupParameter<DataTable>(VariableFrequenciesParameterName, "The relative variable reference frequencies aggregated over the whole population."));
|
|---|
| 141 | Parameters.Add(new LookupParameter<ResultCollection>(ResultsParameterName, "The result collection of the algorithm."));
|
|---|
| 142 | }
|
|---|
| 143 |
|
|---|
| 144 | public override IOperation Apply() {
|
|---|
| 145 | ItemArray<SymbolicExpressionTree> expressions = SymbolicExpressionTreeParameter.ActualValue;
|
|---|
| 146 | ItemArray<DoubleValue> alphas = AlphaParameter.ActualValue;
|
|---|
| 147 | ItemArray<DoubleValue> betas = BetaParameter.ActualValue;
|
|---|
| 148 | var scaledExpressions = from i in Enumerable.Range(0, expressions.Count())
|
|---|
| 149 | let expr = expressions[i]
|
|---|
| 150 | let alpha = alphas[i] == null ? 0.0 : alphas[i].Value
|
|---|
| 151 | let beta = betas[i] == null ? 1.0 : betas[i].Value
|
|---|
| 152 | select new { Expression = expr, Alpha = alpha, Beta = beta };
|
|---|
| 153 | DataAnalysisProblemData problemData = DataAnalysisProblemDataParameter.ActualValue;
|
|---|
| 154 | #region update variable frequencies
|
|---|
| 155 | var inputVariables = problemData.InputVariables.Select(x => x.Value);
|
|---|
| 156 | if (VariableFrequencies == null) {
|
|---|
| 157 | VariableFrequencies = new DataTable("Variable Frequencies", "Relative frequency of variable references aggregated over the whole population.");
|
|---|
| 158 | AddResult("VariableFrequencies", VariableFrequencies);
|
|---|
| 159 | // add a data row for each input variable
|
|---|
| 160 | foreach (var inputVariable in inputVariables)
|
|---|
| 161 | VariableFrequencies.Rows.Add(new DataRow(inputVariable));
|
|---|
| 162 | }
|
|---|
| 163 | foreach (var pair in VariableFrequencyAnalyser.CalculateVariableFrequencies(expressions, inputVariables)) {
|
|---|
| 164 | VariableFrequencies.Rows[pair.Key].Values.Add(pair.Value);
|
|---|
| 165 | }
|
|---|
| 166 | #endregion
|
|---|
| 167 |
|
|---|
| 168 | #region determination of validation-best solution
|
|---|
| 169 | int validationSamplesStart = ValidationSamplesStart.Value;
|
|---|
| 170 | int validationSamplesEnd = ValidationSamplesEnd.Value;
|
|---|
| 171 | var validationValues = problemData.Dataset.GetVariableValues(problemData.TargetVariable.Value, validationSamplesStart, validationSamplesEnd);
|
|---|
| 172 | double upperEstimationLimit = UpperEstimationLimit.Value;
|
|---|
| 173 | double lowerEstimationLimit = LowerEstimationLimit.Value;
|
|---|
| 174 | var currentBestExpression = (from expression in scaledExpressions
|
|---|
| 175 | let validationQuality =
|
|---|
| 176 | SymbolicRegressionScaledMeanSquaredErrorEvaluator.CalculateWithScaling(
|
|---|
| 177 | SymbolicExpressionTreeInterpreter, expression.Expression,
|
|---|
| 178 | lowerEstimationLimit, upperEstimationLimit,
|
|---|
| 179 | problemData.Dataset, problemData.TargetVariable.Value,
|
|---|
| 180 | validationSamplesStart, validationSamplesEnd,
|
|---|
| 181 | expression.Beta, expression.Alpha)
|
|---|
| 182 | select new { Expression = expression, ValidationQuality = validationQuality })
|
|---|
| 183 | .OrderBy(x => x.ValidationQuality)
|
|---|
| 184 | .First();
|
|---|
| 185 |
|
|---|
| 186 | SymbolicRegressionSolution bestOfRunSolution = BestValidationSolutionParameter.ActualValue;
|
|---|
| 187 | #endregion
|
|---|
| 188 | #region update of validation-best solution
|
|---|
| 189 | if (bestOfRunSolution == null) {
|
|---|
| 190 | // no best of run solution yet -> make a solution from the currentBestExpression
|
|---|
| 191 | UpdateBestOfRunSolution(problemData, currentBestExpression.Expression.Expression, SymbolicExpressionTreeInterpreter, lowerEstimationLimit, upperEstimationLimit, currentBestExpression.Expression.Alpha, currentBestExpression.Expression.Beta);
|
|---|
| 192 | } else {
|
|---|
| 193 | // compare quality of current best with best of run solution
|
|---|
| 194 | var estimatedValidationValues = bestOfRunSolution.EstimatedValues.Skip(validationSamplesStart).Take(validationSamplesEnd - validationSamplesStart);
|
|---|
| 195 | var bestOfRunValidationQuality = SimpleMSEEvaluator.Calculate(validationValues, estimatedValidationValues);
|
|---|
| 196 | if (bestOfRunValidationQuality > currentBestExpression.ValidationQuality) {
|
|---|
| 197 | UpdateBestOfRunSolution(problemData, currentBestExpression.Expression.Expression, SymbolicExpressionTreeInterpreter, lowerEstimationLimit, upperEstimationLimit, currentBestExpression.Expression.Alpha, currentBestExpression.Expression.Beta);
|
|---|
| 198 | }
|
|---|
| 199 | }
|
|---|
| 200 | #endregion
|
|---|
| 201 | return base.Apply();
|
|---|
| 202 | }
|
|---|
| 203 |
|
|---|
| 204 | private void UpdateBestOfRunSolution(DataAnalysisProblemData problemData, SymbolicExpressionTree tree, ISymbolicExpressionTreeInterpreter interpreter,
|
|---|
| 205 | double lowerEstimationLimit, double upperEstimationLimit,
|
|---|
| 206 | double alpha, double beta) {
|
|---|
| 207 | var newBestSolution = CreateDataAnalysisSolution(problemData, tree, interpreter, lowerEstimationLimit, upperEstimationLimit, alpha, beta);
|
|---|
| 208 | if (BestValidationSolutionParameter.ActualValue == null)
|
|---|
| 209 | BestValidationSolutionParameter.ActualValue = newBestSolution;
|
|---|
| 210 | else
|
|---|
| 211 | // only update model
|
|---|
| 212 | BestValidationSolutionParameter.ActualValue.Model = newBestSolution.Model;
|
|---|
| 213 |
|
|---|
| 214 | AddResult("NumberOfInputVariables", new IntValue(CountInputVariables(tree)));
|
|---|
| 215 |
|
|---|
| 216 | var trainingValues = problemData.Dataset.GetVariableValues(problemData.TargetVariable.Value, problemData.TrainingSamplesStart.Value, problemData.TrainingSamplesEnd.Value);
|
|---|
| 217 | var testValues = problemData.Dataset.GetVariableValues(problemData.TargetVariable.Value, problemData.TestSamplesStart.Value, problemData.TestSamplesEnd.Value);
|
|---|
| 218 |
|
|---|
| 219 | AddResult("MeanSquaredError (Training)", new DoubleValue(SimpleMSEEvaluator.Calculate(trainingValues, newBestSolution.EstimatedTrainingValues)));
|
|---|
| 220 | AddResult("MeanRelativeError (Training)", new PercentValue(SimpleMeanAbsolutePercentageErrorEvaluator.Calculate(trainingValues, newBestSolution.EstimatedTrainingValues)));
|
|---|
| 221 | AddResult("RSquared (Training)", new DoubleValue(SimpleRSquaredEvaluator.Calculate(trainingValues, newBestSolution.EstimatedTrainingValues)));
|
|---|
| 222 |
|
|---|
| 223 | AddResult("MeanSquaredError (Test)", new DoubleValue(SimpleMSEEvaluator.Calculate(testValues, newBestSolution.EstimatedTestValues)));
|
|---|
| 224 | AddResult("MeanRelativeError (Test)", new PercentValue(SimpleMeanAbsolutePercentageErrorEvaluator.Calculate(testValues, newBestSolution.EstimatedTestValues)));
|
|---|
| 225 | AddResult("RSquared (Test)", new DoubleValue(SimpleRSquaredEvaluator.Calculate(testValues, newBestSolution.EstimatedTestValues)));
|
|---|
| 226 | }
|
|---|
| 227 |
|
|---|
| 228 | private int CountInputVariables(SymbolicExpressionTree tree) {
|
|---|
| 229 | return (from node in tree.IterateNodesPrefix().OfType<VariableTreeNode>()
|
|---|
| 230 | select node.VariableName)
|
|---|
| 231 | .Distinct()
|
|---|
| 232 | .Count();
|
|---|
| 233 | }
|
|---|
| 234 |
|
|---|
| 235 | private void AddResult(string resultName, IItem value) {
|
|---|
| 236 | var resultCollection = ResultParameter.ActualValue;
|
|---|
| 237 | if (resultCollection.ContainsKey(resultName)) {
|
|---|
| 238 | resultCollection[resultName].Value = value;
|
|---|
| 239 | } else {
|
|---|
| 240 | resultCollection.Add(new Result(resultName, value));
|
|---|
| 241 | }
|
|---|
| 242 | }
|
|---|
| 243 |
|
|---|
| 244 | private SymbolicRegressionSolution CreateDataAnalysisSolution(DataAnalysisProblemData problemData, SymbolicExpressionTree tree, ISymbolicExpressionTreeInterpreter interpreter,
|
|---|
| 245 | double lowerEstimationLimit, double upperEstimationLimit,
|
|---|
| 246 | double alpha, double beta) {
|
|---|
| 247 | var mainBranch = tree.Root.SubTrees[0].SubTrees[0];
|
|---|
| 248 | var scaledMainBranch = MakeSum(MakeProduct(beta, mainBranch), alpha);
|
|---|
| 249 |
|
|---|
| 250 | // remove the main branch before cloning to prevent cloning of sub-trees
|
|---|
| 251 | tree.Root.SubTrees[0].RemoveSubTree(0);
|
|---|
| 252 | var scaledTree = (SymbolicExpressionTree)tree.Clone();
|
|---|
| 253 | // insert main branch into the original tree again
|
|---|
| 254 | tree.Root.SubTrees[0].InsertSubTree(0, mainBranch);
|
|---|
| 255 | // insert the scaled main branch into the cloned tree
|
|---|
| 256 | scaledTree.Root.SubTrees[0].InsertSubTree(0, scaledMainBranch);
|
|---|
| 257 | // create a new solution using the scaled tree
|
|---|
| 258 | var model = new SymbolicRegressionModel(interpreter, scaledTree, problemData.InputVariables.Select(s => s.Value));
|
|---|
| 259 | return new SymbolicRegressionSolution(problemData, model, lowerEstimationLimit, upperEstimationLimit);
|
|---|
| 260 | }
|
|---|
| 261 |
|
|---|
| 262 | private SymbolicExpressionTreeNode MakeSum(SymbolicExpressionTreeNode treeNode, double alpha) {
|
|---|
| 263 | var node = (new Addition()).CreateTreeNode();
|
|---|
| 264 | var alphaConst = MakeConstant(alpha);
|
|---|
| 265 | node.AddSubTree(treeNode);
|
|---|
| 266 | node.AddSubTree(alphaConst);
|
|---|
| 267 | return node;
|
|---|
| 268 | }
|
|---|
| 269 |
|
|---|
| 270 | private SymbolicExpressionTreeNode MakeProduct(double beta, SymbolicExpressionTreeNode treeNode) {
|
|---|
| 271 | var node = (new Multiplication()).CreateTreeNode();
|
|---|
| 272 | var betaConst = MakeConstant(beta);
|
|---|
| 273 | node.AddSubTree(treeNode);
|
|---|
| 274 | node.AddSubTree(betaConst);
|
|---|
| 275 | return node;
|
|---|
| 276 | }
|
|---|
| 277 |
|
|---|
| 278 | private SymbolicExpressionTreeNode MakeConstant(double c) {
|
|---|
| 279 | var node = (ConstantTreeNode)(new Constant()).CreateTreeNode();
|
|---|
| 280 | node.Value = c;
|
|---|
| 281 | return node;
|
|---|
| 282 | }
|
|---|
| 283 | }
|
|---|
| 284 | }
|
|---|
