| 1 | #region License Information
|
|---|
| 2 | /* HeuristicLab
|
|---|
| 3 | * Copyright (C) 2002-2012 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 HeuristicLab.Common;
|
|---|
| 24 | using HeuristicLab.Data;
|
|---|
| 25 | using HeuristicLab.Optimization;
|
|---|
| 26 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
|---|
| 27 |
|
|---|
| 28 | namespace HeuristicLab.Problems.DataAnalysis {
|
|---|
| 29 | [StorableClass]
|
|---|
| 30 | public abstract class RegressionSolutionBase : DataAnalysisSolution, IRegressionSolution {
|
|---|
| 31 | protected const string TrainingMeanSquaredErrorResultName = "Mean squared error (training)";
|
|---|
| 32 | protected const string TestMeanSquaredErrorResultName = "Mean squared error (test)";
|
|---|
| 33 | protected const string TrainingMeanAbsoluteErrorResultName = "Mean absolute error (training)";
|
|---|
| 34 | protected const string TestMeanAbsoluteErrorResultName = "Mean absolute error (test)";
|
|---|
| 35 | protected const string TrainingSquaredCorrelationResultName = "Pearson's R² (training)";
|
|---|
| 36 | protected const string TestSquaredCorrelationResultName = "Pearson's R² (test)";
|
|---|
| 37 | protected const string TrainingRelativeErrorResultName = "Average relative error (training)";
|
|---|
| 38 | protected const string TestRelativeErrorResultName = "Average relative error (test)";
|
|---|
| 39 | protected const string TrainingNormalizedMeanSquaredErrorResultName = "Normalized mean squared error (training)";
|
|---|
| 40 | protected const string TestNormalizedMeanSquaredErrorResultName = "Normalized mean squared error (test)";
|
|---|
| 41 | protected const string TrainingMeanErrorResultName = "Mean error (training)";
|
|---|
| 42 | protected const string TestMeanErrorResultName = "Mean error (test)";
|
|---|
| 43 |
|
|---|
| 44 | protected const string TrainingMeanSquaredErrorResultDescription = "Mean of squared errors of the model on the training partition";
|
|---|
| 45 | protected const string TestMeanSquaredErrorResultDescription = "Mean of squared errors of the model on the test partition";
|
|---|
| 46 | protected const string TrainingMeanAbsoluteErrorResultDescription = "Mean of absolute errors of the model on the training partition";
|
|---|
| 47 | protected const string TestMeanAbsoluteErrorResultDescription = "Mean of absolute errors of the model on the test partition";
|
|---|
| 48 | protected const string TrainingSquaredCorrelationResultDescription = "Squared Pearson's correlation coefficient of the model output and the actual values on the training partition";
|
|---|
| 49 | protected const string TestSquaredCorrelationResultDescription = "Squared Pearson's correlation coefficient of the model output and the actual values on the test partition";
|
|---|
| 50 | protected const string TrainingRelativeErrorResultDescription = "Average of the relative errors of the model output and the actual values on the training partition";
|
|---|
| 51 | protected const string TestRelativeErrorResultDescription = "Average of the relative errors of the model output and the actual values on the test partition";
|
|---|
| 52 | protected const string TrainingNormalizedMeanSquaredErrorResultDescription = "Normalized mean of squared errors of the model on the training partition";
|
|---|
| 53 | protected const string TestNormalizedMeanSquaredErrorResultDescription = "Normalized mean of squared errors of the model on the test partition";
|
|---|
| 54 | protected const string TrainingMeanErrorResultDescription = "Mean of errors of the model on the training partition";
|
|---|
| 55 | protected const string TestMeanErrorResultDescription = "Mean of errors of the model on the test partition";
|
|---|
| 56 |
|
|---|
| 57 | public new IRegressionModel Model {
|
|---|
| 58 | get { return (IRegressionModel)base.Model; }
|
|---|
| 59 | protected set { base.Model = value; }
|
|---|
| 60 | }
|
|---|
| 61 |
|
|---|
| 62 | public new IRegressionProblemData ProblemData {
|
|---|
| 63 | get { return (IRegressionProblemData)base.ProblemData; }
|
|---|
| 64 | set { base.ProblemData = value; }
|
|---|
| 65 | }
|
|---|
| 66 |
|
|---|
| 67 | public abstract IEnumerable<double> EstimatedValues { get; }
|
|---|
| 68 | public abstract IEnumerable<double> EstimatedTrainingValues { get; }
|
|---|
| 69 | public abstract IEnumerable<double> EstimatedTestValues { get; }
|
|---|
| 70 | public abstract IEnumerable<double> GetEstimatedValues(IEnumerable<int> rows);
|
|---|
| 71 |
|
|---|
| 72 | #region Results
|
|---|
| 73 | public double TrainingMeanSquaredError {
|
|---|
| 74 | get { return ((DoubleValue)this[TrainingMeanSquaredErrorResultName].Value).Value; }
|
|---|
| 75 | private set { ((DoubleValue)this[TrainingMeanSquaredErrorResultName].Value).Value = value; }
|
|---|
| 76 | }
|
|---|
| 77 | public double TestMeanSquaredError {
|
|---|
| 78 | get { return ((DoubleValue)this[TestMeanSquaredErrorResultName].Value).Value; }
|
|---|
| 79 | private set { ((DoubleValue)this[TestMeanSquaredErrorResultName].Value).Value = value; }
|
|---|
| 80 | }
|
|---|
| 81 | public double TrainingMeanAbsoluteError {
|
|---|
| 82 | get { return ((DoubleValue)this[TrainingMeanAbsoluteErrorResultName].Value).Value; }
|
|---|
| 83 | private set { ((DoubleValue)this[TrainingMeanAbsoluteErrorResultName].Value).Value = value; }
|
|---|
| 84 | }
|
|---|
| 85 | public double TestMeanAbsoluteError {
|
|---|
| 86 | get { return ((DoubleValue)this[TestMeanAbsoluteErrorResultName].Value).Value; }
|
|---|
| 87 | private set { ((DoubleValue)this[TestMeanAbsoluteErrorResultName].Value).Value = value; }
|
|---|
| 88 | }
|
|---|
| 89 | public double TrainingRSquared {
|
|---|
| 90 | get { return ((DoubleValue)this[TrainingSquaredCorrelationResultName].Value).Value; }
|
|---|
| 91 | private set { ((DoubleValue)this[TrainingSquaredCorrelationResultName].Value).Value = value; }
|
|---|
| 92 | }
|
|---|
| 93 | public double TestRSquared {
|
|---|
| 94 | get { return ((DoubleValue)this[TestSquaredCorrelationResultName].Value).Value; }
|
|---|
| 95 | private set { ((DoubleValue)this[TestSquaredCorrelationResultName].Value).Value = value; }
|
|---|
| 96 | }
|
|---|
| 97 | public double TrainingRelativeError {
|
|---|
| 98 | get { return ((DoubleValue)this[TrainingRelativeErrorResultName].Value).Value; }
|
|---|
| 99 | private set { ((DoubleValue)this[TrainingRelativeErrorResultName].Value).Value = value; }
|
|---|
| 100 | }
|
|---|
| 101 | public double TestRelativeError {
|
|---|
| 102 | get { return ((DoubleValue)this[TestRelativeErrorResultName].Value).Value; }
|
|---|
| 103 | private set { ((DoubleValue)this[TestRelativeErrorResultName].Value).Value = value; }
|
|---|
| 104 | }
|
|---|
| 105 | public double TrainingNormalizedMeanSquaredError {
|
|---|
| 106 | get { return ((DoubleValue)this[TrainingNormalizedMeanSquaredErrorResultName].Value).Value; }
|
|---|
| 107 | private set { ((DoubleValue)this[TrainingNormalizedMeanSquaredErrorResultName].Value).Value = value; }
|
|---|
| 108 | }
|
|---|
| 109 | public double TestNormalizedMeanSquaredError {
|
|---|
| 110 | get { return ((DoubleValue)this[TestNormalizedMeanSquaredErrorResultName].Value).Value; }
|
|---|
| 111 | private set { ((DoubleValue)this[TestNormalizedMeanSquaredErrorResultName].Value).Value = value; }
|
|---|
| 112 | }
|
|---|
| 113 | public double TrainingMeanError {
|
|---|
| 114 | get { return ((DoubleValue)this[TrainingMeanErrorResultName].Value).Value; }
|
|---|
| 115 | private set { ((DoubleValue)this[TrainingMeanErrorResultName].Value).Value = value; }
|
|---|
| 116 | }
|
|---|
| 117 | public double TestMeanError {
|
|---|
| 118 | get { return ((DoubleValue)this[TestMeanErrorResultName].Value).Value; }
|
|---|
| 119 | private set { ((DoubleValue)this[TestMeanErrorResultName].Value).Value = value; }
|
|---|
| 120 | }
|
|---|
| 121 | #endregion
|
|---|
| 122 |
|
|---|
| 123 | [StorableConstructor]
|
|---|
| 124 | protected RegressionSolutionBase(bool deserializing) : base(deserializing) { }
|
|---|
| 125 | protected RegressionSolutionBase(RegressionSolutionBase original, Cloner cloner)
|
|---|
| 126 | : base(original, cloner) {
|
|---|
| 127 | }
|
|---|
| 128 | protected RegressionSolutionBase(IRegressionModel model, IRegressionProblemData problemData)
|
|---|
| 129 | : base(model, problemData) {
|
|---|
| 130 | Add(new Result(TrainingMeanSquaredErrorResultName, TrainingMeanSquaredErrorResultDescription, new DoubleValue()));
|
|---|
| 131 | Add(new Result(TestMeanSquaredErrorResultName, TestMeanSquaredErrorResultDescription, new DoubleValue()));
|
|---|
| 132 | Add(new Result(TrainingMeanAbsoluteErrorResultName, TrainingMeanAbsoluteErrorResultDescription, new DoubleValue()));
|
|---|
| 133 | Add(new Result(TestMeanAbsoluteErrorResultName, TestMeanAbsoluteErrorResultDescription, new DoubleValue()));
|
|---|
| 134 | Add(new Result(TrainingSquaredCorrelationResultName, TrainingSquaredCorrelationResultDescription, new DoubleValue()));
|
|---|
| 135 | Add(new Result(TestSquaredCorrelationResultName, TestSquaredCorrelationResultDescription, new DoubleValue()));
|
|---|
| 136 | Add(new Result(TrainingRelativeErrorResultName, TrainingRelativeErrorResultDescription, new PercentValue()));
|
|---|
| 137 | Add(new Result(TestRelativeErrorResultName, TestRelativeErrorResultDescription, new PercentValue()));
|
|---|
| 138 | Add(new Result(TrainingNormalizedMeanSquaredErrorResultName, TrainingNormalizedMeanSquaredErrorResultDescription, new DoubleValue()));
|
|---|
| 139 | Add(new Result(TestNormalizedMeanSquaredErrorResultName, TestNormalizedMeanSquaredErrorResultDescription, new DoubleValue()));
|
|---|
| 140 | Add(new Result(TrainingMeanErrorResultName, TrainingMeanErrorResultDescription, new DoubleValue()));
|
|---|
| 141 | Add(new Result(TestMeanErrorResultName, TestMeanErrorResultDescription, new DoubleValue()));
|
|---|
| 142 | }
|
|---|
| 143 |
|
|---|
| 144 | [StorableHook(HookType.AfterDeserialization)]
|
|---|
| 145 | private void AfterDeserialization() {
|
|---|
| 146 | // BackwardsCompatibility3.4
|
|---|
| 147 |
|
|---|
| 148 | #region Backwards compatible code, remove with 3.5
|
|---|
| 149 |
|
|---|
| 150 | if (!ContainsKey(TrainingMeanAbsoluteErrorResultName)) {
|
|---|
| 151 | OnlineCalculatorError errorState;
|
|---|
| 152 | Add(new Result(TrainingMeanAbsoluteErrorResultName, "Mean of absolute errors of the model on the training partition", new DoubleValue()));
|
|---|
| 153 | double trainingMAE = OnlineMeanAbsoluteErrorCalculator.Calculate(EstimatedTrainingValues, ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TrainingIndices), out errorState);
|
|---|
| 154 | TrainingMeanAbsoluteError = errorState == OnlineCalculatorError.None ? trainingMAE : double.NaN;
|
|---|
| 155 | }
|
|---|
| 156 |
|
|---|
| 157 | if (!ContainsKey(TestMeanAbsoluteErrorResultName)) {
|
|---|
| 158 | OnlineCalculatorError errorState;
|
|---|
| 159 | Add(new Result(TestMeanAbsoluteErrorResultName, "Mean of absolute errors of the model on the test partition", new DoubleValue()));
|
|---|
| 160 | double testMAE = OnlineMeanAbsoluteErrorCalculator.Calculate(EstimatedTestValues, ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TestIndices), out errorState);
|
|---|
| 161 | TestMeanAbsoluteError = errorState == OnlineCalculatorError.None ? testMAE : double.NaN;
|
|---|
| 162 | }
|
|---|
| 163 |
|
|---|
| 164 | if (!ContainsKey(TrainingMeanErrorResultName)) {
|
|---|
| 165 | OnlineCalculatorError errorState;
|
|---|
| 166 | Add(new Result(TrainingMeanErrorResultName, "Mean of errors of the model on the training partition", new DoubleValue()));
|
|---|
| 167 | double trainingME = OnlineMeanErrorCalculator.Calculate(EstimatedTrainingValues, ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TrainingIndices), out errorState);
|
|---|
| 168 | TrainingMeanError = errorState == OnlineCalculatorError.None ? trainingME : double.NaN;
|
|---|
| 169 | }
|
|---|
| 170 | if (!ContainsKey(TestMeanErrorResultName)) {
|
|---|
| 171 | OnlineCalculatorError errorState;
|
|---|
| 172 | Add(new Result(TestMeanErrorResultName, "Mean of errors of the model on the test partition", new DoubleValue()));
|
|---|
| 173 | double testME = OnlineMeanErrorCalculator.Calculate(EstimatedTestValues, ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TestIndices), out errorState);
|
|---|
| 174 | TestMeanError = errorState == OnlineCalculatorError.None ? testME : double.NaN;
|
|---|
| 175 | }
|
|---|
| 176 | #endregion
|
|---|
| 177 | }
|
|---|
| 178 |
|
|---|
| 179 | protected override void RecalculateResults() {
|
|---|
| 180 | CalculateRegressionResults();
|
|---|
| 181 | }
|
|---|
| 182 |
|
|---|
| 183 | protected void CalculateRegressionResults() {
|
|---|
| 184 | IEnumerable<double> estimatedTrainingValues = EstimatedTrainingValues; // cache values
|
|---|
| 185 | IEnumerable<double> originalTrainingValues = ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TrainingIndices);
|
|---|
| 186 | IEnumerable<double> estimatedTestValues = EstimatedTestValues; // cache values
|
|---|
| 187 | IEnumerable<double> originalTestValues = ProblemData.Dataset.GetDoubleValues(ProblemData.TargetVariable, ProblemData.TestIndices);
|
|---|
| 188 |
|
|---|
| 189 | OnlineCalculatorError errorState;
|
|---|
| 190 | double trainingMSE = OnlineMeanSquaredErrorCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 191 | TrainingMeanSquaredError = errorState == OnlineCalculatorError.None ? trainingMSE : double.NaN;
|
|---|
| 192 | double testMSE = OnlineMeanSquaredErrorCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 193 | TestMeanSquaredError = errorState == OnlineCalculatorError.None ? testMSE : double.NaN;
|
|---|
| 194 |
|
|---|
| 195 | double trainingMAE = OnlineMeanAbsoluteErrorCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 196 | TrainingMeanAbsoluteError = errorState == OnlineCalculatorError.None ? trainingMAE : double.NaN;
|
|---|
| 197 | double testMAE = OnlineMeanAbsoluteErrorCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 198 | TestMeanAbsoluteError = errorState == OnlineCalculatorError.None ? testMAE : double.NaN;
|
|---|
| 199 |
|
|---|
| 200 | double trainingR2 = OnlinePearsonsRSquaredCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 201 | TrainingRSquared = errorState == OnlineCalculatorError.None ? trainingR2 : double.NaN;
|
|---|
| 202 | double testR2 = OnlinePearsonsRSquaredCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 203 | TestRSquared = errorState == OnlineCalculatorError.None ? testR2 : double.NaN;
|
|---|
| 204 |
|
|---|
| 205 | double trainingRelError = OnlineMeanAbsolutePercentageErrorCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 206 | TrainingRelativeError = errorState == OnlineCalculatorError.None ? trainingRelError : double.NaN;
|
|---|
| 207 | double testRelError = OnlineMeanAbsolutePercentageErrorCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 208 | TestRelativeError = errorState == OnlineCalculatorError.None ? testRelError : double.NaN;
|
|---|
| 209 |
|
|---|
| 210 | double trainingNMSE = OnlineNormalizedMeanSquaredErrorCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 211 | TrainingNormalizedMeanSquaredError = errorState == OnlineCalculatorError.None ? trainingNMSE : double.NaN;
|
|---|
| 212 | double testNMSE = OnlineNormalizedMeanSquaredErrorCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 213 | TestNormalizedMeanSquaredError = errorState == OnlineCalculatorError.None ? testNMSE : double.NaN;
|
|---|
| 214 |
|
|---|
| 215 | double trainingME = OnlineMeanErrorCalculator.Calculate(originalTrainingValues, estimatedTrainingValues, out errorState);
|
|---|
| 216 | TrainingMeanError = errorState == OnlineCalculatorError.None ? trainingME : double.NaN;
|
|---|
| 217 | double testME = OnlineMeanErrorCalculator.Calculate(originalTestValues, estimatedTestValues, out errorState);
|
|---|
| 218 | TestMeanError = errorState == OnlineCalculatorError.None ? testME : double.NaN;
|
|---|
| 219 | }
|
|---|
| 220 | }
|
|---|
| 221 | }
|
|---|
