source: branches/SensitivityEvaluator/HeuristicLab.Problems.DataAnalysis.Symbolic.Classification/3.4/SingleObjective/SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator.cs @ 12449

Last change on this file since 12449 was 12449, checked in by ehopf, 5 years ago

#2361: Changed the calculation of the error in the OnlineWeightedResidualsMeanSquaredErrorCalculator. The weight is now only applied once to the error.

File size: 8.9 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2015 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.Core;
27using HeuristicLab.Data;
28using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
29using HeuristicLab.Parameters;
30using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
31
32namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Classification {
33  [Item("Weighted Residuals Mean Squared Error Evaluator", @"A modified mean squared error evaluator that enables the possibility to weight residuals differently.
34The first residual category belongs to estimated values which definitely belong to a specific class because the estimated value is located above the maximum or below the minimum of all the class values (DefiniteResidualsWeight).
35The second residual category represents residuals which belong to the positive class whereby the estimated value is located between the positive and a negative class (PositiveClassResidualsWeight).
36All other cases are represented by the third category (NegativeClassesResidualsWeight).
37The weight gets multiplied to the squared error. Note that the Evaluator acts like a normal MSE-Evaluator if all the weights are set to 1.")]
38  [StorableClass]
39  public class SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator : SymbolicClassificationSingleObjectiveEvaluator {
40    private const string DefiniteResidualsWeightParameterName = "DefiniteResidualsWeight";
41    private const string PositiveClassResidualsWeightParameterName = "PositiveClassResidualsWeight";
42    private const string NegativeClassesResidualsWeightParameterName = "NegativeClassesResidualsWeight";
43    [StorableConstructor]
44    protected SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator(bool deserializing) : base(deserializing) { }
45    protected SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator(SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator original, Cloner cloner)
46      : base(original, cloner) {
47    }
48    public override IDeepCloneable Clone(Cloner cloner) {
49      return new SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator(this, cloner);
50    }
51
52    public SymbolicClassificationSingleObjectiveWeightedResidualsMeanSquaredErrorEvaluator()
53      : base() {
54      Parameters.Add(new FixedValueParameter<DoubleValue>(DefiniteResidualsWeightParameterName, "Weight of residuals which definitely belong to a specific class because the estimated values is located above the maximum or below the minimum of all the class values.", new DoubleValue(1)));
55      Parameters.Add(new FixedValueParameter<DoubleValue>(PositiveClassResidualsWeightParameterName, "Weight of residuals which belong to the positive class whereby the estimated value is located between the positive and a negative class.", new DoubleValue(1)));
56      Parameters.Add(new FixedValueParameter<DoubleValue>(NegativeClassesResidualsWeightParameterName, "Weight of residuals which are not covered by the DefiniteResidualsWeight or the PositiveClassResidualsWeight.", new DoubleValue(1)));
57    }
58
59    #region parameter properties
60    public IFixedValueParameter<DoubleValue> DefiniteResidualsWeightParameter {
61      get { return (IFixedValueParameter<DoubleValue>)Parameters[DefiniteResidualsWeightParameterName]; }
62    }
63    public IFixedValueParameter<DoubleValue> PositiveClassResidualsWeightParameter {
64      get { return (IFixedValueParameter<DoubleValue>)Parameters[PositiveClassResidualsWeightParameterName]; }
65    }
66    public IFixedValueParameter<DoubleValue> NegativeClassesResidualsWeightParameter {
67      get { return (IFixedValueParameter<DoubleValue>)Parameters[NegativeClassesResidualsWeightParameterName]; }
68    }
69    #endregion
70
71    #region properties
72    public override bool Maximization { get { return false; } }
73
74    public double DefiniteResidualsWeight {
75      get {
76        return DefiniteResidualsWeightParameter.Value.Value;
77      }
78    }
79    public double PositiveClassResidualsWeight {
80      get {
81        return PositiveClassResidualsWeightParameter.Value.Value;
82      }
83    }
84    public double NegativeClassesResidualsWeight {
85      get {
86        return NegativeClassesResidualsWeightParameter.Value.Value;
87      }
88    }
89    #endregion
90
91    public override IOperation InstrumentedApply() {
92      IEnumerable<int> rows = GenerateRowsToEvaluate();
93      var solution = SymbolicExpressionTreeParameter.ActualValue;
94      double quality = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, solution, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, ProblemDataParameter.ActualValue, rows, ApplyLinearScalingParameter.ActualValue.Value,
95        DefiniteResidualsWeight, PositiveClassResidualsWeight, NegativeClassesResidualsWeight);
96      QualityParameter.ActualValue = new DoubleValue(quality);
97      return base.InstrumentedApply();
98    }
99
100    public static double Calculate(ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, ISymbolicExpressionTree solution, double lowerEstimationLimit, double upperEstimationLimit, IClassificationProblemData problemData, IEnumerable<int> rows, bool applyLinearScaling,
101      double definiteResidualsWeight, double positiveClassResidualsWeight, double negativeClassesResidualsWeight) {
102      IEnumerable<double> estimatedValues = interpreter.GetSymbolicExpressionTreeValues(solution, problemData.Dataset, rows);
103      IEnumerable<double> targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
104      OnlineCalculatorError errorState;
105
106      double positiveClassValue = problemData.GetClassValue(problemData.PositiveClass);
107      //get class values min/max
108      double classValuesMin = problemData.ClassValues.ElementAtOrDefault(0);
109      double classValuesMax = classValuesMin;
110      foreach (double classValue in problemData.ClassValues) {
111        if (classValuesMin > classValue) classValuesMin = classValue;
112        if (classValuesMax < classValue) classValuesMax = classValue;
113      }
114
115      double quality;
116      if (applyLinearScaling) {
117        var calculator = new OnlineWeightedResidualsMeanSquaredErrorCalculator(positiveClassValue, classValuesMax, classValuesMin,
118          definiteResidualsWeight, positiveClassResidualsWeight, negativeClassesResidualsWeight);
119        CalculateWithScaling(targetValues, estimatedValues, lowerEstimationLimit, upperEstimationLimit, calculator, problemData.Dataset.Rows);
120        errorState = calculator.ErrorState;
121        quality = calculator.WeightedResidualsMeanSquaredError;
122      } else {
123        IEnumerable<double> boundedEstimatedValues = estimatedValues.LimitToRange(lowerEstimationLimit, upperEstimationLimit);
124        quality = OnlineWeightedResidualsMeanSquaredErrorCalculator.Calculate(targetValues, boundedEstimatedValues, positiveClassValue, classValuesMax,
125          classValuesMin, definiteResidualsWeight, positiveClassResidualsWeight, negativeClassesResidualsWeight, out errorState);
126      }
127      if (errorState != OnlineCalculatorError.None) return Double.NaN;
128      return quality;
129    }
130
131    public override double Evaluate(IExecutionContext context, ISymbolicExpressionTree tree, IClassificationProblemData problemData, IEnumerable<int> rows) {
132      SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = context;
133      EstimationLimitsParameter.ExecutionContext = context;
134      ApplyLinearScalingParameter.ExecutionContext = context;
135
136      double quality = Calculate(SymbolicDataAnalysisTreeInterpreterParameter.ActualValue, tree, EstimationLimitsParameter.ActualValue.Lower, EstimationLimitsParameter.ActualValue.Upper, problemData, rows, ApplyLinearScalingParameter.ActualValue.Value, DefiniteResidualsWeight, PositiveClassResidualsWeight, NegativeClassesResidualsWeight);
137
138      SymbolicDataAnalysisTreeInterpreterParameter.ExecutionContext = null;
139      EstimationLimitsParameter.ExecutionContext = null;
140      ApplyLinearScalingParameter.ExecutionContext = null;
141
142      return quality;
143    }
144  }
145}
Note: See TracBrowser for help on using the repository browser.