Free cookie consent management tool by TermsFeed Policy Generator

source: trunk/sources/HeuristicLab.StructureIdentification/Evaluation/TheilInequalityCoefficientEvaluator.cs @ 395

Last change on this file since 395 was 395, checked in by gkronber, 16 years ago

added evaluator for mean absolute percentage error and added a parameter for the evaluator for theil's inequality to determine whether to calculate the coefficient for the change or for the absolute value

File size: 3.6 KB
Line 
1#region License Information
2/* HeuristicLab
3 * Copyright (C) 2002-2008 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 System.Text;
26using HeuristicLab.Core;
27using HeuristicLab.Data;
28using HeuristicLab.Operators;
29using HeuristicLab.Functions;
30using HeuristicLab.DataAnalysis;
31
32namespace HeuristicLab.StructureIdentification {
33  public class TheilInequalityCoefficientEvaluator : GPEvaluatorBase {
34    public override string Description {
35      get {
36        return @"Evaluates 'FunctionTree' for all samples of 'Dataset' and calculates
37the 'Theil inequality coefficient (scale invariant)' of estimated values vs. real values of 'TargetVariable'.";
38      }
39    }
40
41    public TheilInequalityCoefficientEvaluator()
42      : base() {
43      AddVariableInfo(new VariableInfo("Differential", "Wether to calculate the coefficient for the predicted change vs. original change or for the absolute prediction vs. original value", typeof(BoolData), VariableKind.In));
44    }
45
46    public override double Evaluate(IScope scope, IFunctionTree functionTree, int targetVariable, Dataset dataset) {
47      int trainingStart = GetVariableValue<IntData>("TrainingSamplesStart", scope, true).Data;
48      int trainingEnd = GetVariableValue<IntData>("TrainingSamplesEnd", scope, true).Data;
49      bool difference = GetVariableValue<BoolData>("Differential", scope, true).Data;
50      double errorsSquaredSum = 0.0;
51      double estimatedSquaredSum = 0.0;
52      double originalSquaredSum = 0.0;
53      functionTree.PrepareEvaluation(dataset);
54      for(int sample = trainingStart; sample < trainingEnd; sample++) {
55        double prevValue = 0.0;
56        if(difference) prevValue = dataset.GetValue(sample - 1, targetVariable);
57        double estimatedChange = functionTree.Evaluate(sample) - prevValue;
58        double originalChange = dataset.GetValue(sample, targetVariable) - prevValue;
59        if(!double.IsNaN(originalChange) && !double.IsInfinity(originalChange)) {
60          if(double.IsNaN(estimatedChange) || double.IsInfinity(estimatedChange))
61            estimatedChange = maximumPunishment;
62          else if(estimatedChange > maximumPunishment)
63            estimatedChange = maximumPunishment;
64          else if(estimatedChange < -maximumPunishment)
65            estimatedChange = - maximumPunishment;
66
67          double error = estimatedChange - originalChange;
68          errorsSquaredSum += error * error;
69          estimatedSquaredSum += estimatedChange * estimatedChange;
70          originalSquaredSum += originalChange * originalChange;
71        }
72      }
73      int nSamples = trainingEnd - trainingStart;
74      double quality = Math.Sqrt(errorsSquaredSum / nSamples) / (Math.Sqrt(estimatedSquaredSum/nSamples) + Math.Sqrt(originalSquaredSum/nSamples));
75      if(double.IsNaN(quality) || double.IsInfinity(quality))
76        quality = double.MaxValue;
77      return quality;
78    }
79  }
80}
Note: See TracBrowser for help on using the repository browser.