[11315] | 1 | #region License Information
|
---|
| 2 |
|
---|
| 3 | /* HeuristicLab
|
---|
[17181] | 4 | * Copyright (C) Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[11315] | 5 | *
|
---|
| 6 | * This file is part of HeuristicLab.
|
---|
| 7 | *
|
---|
| 8 | * HeuristicLab is free software: you can redistribute it and/or modify
|
---|
| 9 | * it under the terms of the GNU General Public License as published by
|
---|
| 10 | * the Free Software Foundation, either version 3 of the License, or
|
---|
| 11 | * (at your option) any later version.
|
---|
| 12 | *
|
---|
| 13 | * HeuristicLab is distributed in the hope that it will be useful,
|
---|
| 14 | * but WITHOUT ANY WARRANTY; without even the implied warranty of
|
---|
| 15 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
---|
| 16 | * GNU General Public License for more details.
|
---|
| 17 | *
|
---|
| 18 | * You should have received a copy of the GNU General Public License
|
---|
| 19 | * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
|
---|
| 20 | */
|
---|
| 21 |
|
---|
| 22 | #endregion
|
---|
| 23 |
|
---|
| 24 | using System;
|
---|
| 25 | using System.Collections.Generic;
|
---|
| 26 | using System.Linq;
|
---|
| 27 | using System.Linq.Expressions;
|
---|
| 28 | using System.Threading.Tasks;
|
---|
[17157] | 29 | using HEAL.Attic;
|
---|
[11315] | 30 | using HeuristicLab.Common;
|
---|
[11901] | 31 | using HeuristicLab.Core;
|
---|
[11315] | 32 | using HeuristicLab.Data;
|
---|
[11901] | 33 | using HeuristicLab.Parameters;
|
---|
[11315] | 34 | using HeuristicLab.Problems.DataAnalysis;
|
---|
[11901] | 35 | using HeuristicLab.Random;
|
---|
[11315] | 36 |
|
---|
| 37 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
[11901] | 38 | [Item("RFParameter", "A random forest parameter collection")]
|
---|
[17097] | 39 | [StorableType("40E482DA-63C5-4D39-97C7-63701CF1D021")]
|
---|
[11901] | 40 | public class RFParameter : ParameterCollection {
|
---|
| 41 | public RFParameter() {
|
---|
| 42 | base.Add(new FixedValueParameter<IntValue>("N", "The number of random forest trees", new IntValue(50)));
|
---|
| 43 | base.Add(new FixedValueParameter<DoubleValue>("M", "The ratio of features that will be used in the construction of individual trees (0<m<=1)", new DoubleValue(0.1)));
|
---|
| 44 | base.Add(new FixedValueParameter<DoubleValue>("R", "The ratio of the training set that will be used in the construction of individual trees (0<r<=1)", new DoubleValue(0.1)));
|
---|
| 45 | }
|
---|
[11315] | 46 |
|
---|
[11901] | 47 | [StorableConstructor]
|
---|
[17097] | 48 | protected RFParameter(StorableConstructorFlag _) : base(_) {
|
---|
[11901] | 49 | }
|
---|
[11315] | 50 |
|
---|
[11901] | 51 | protected RFParameter(RFParameter original, Cloner cloner)
|
---|
| 52 | : base(original, cloner) {
|
---|
| 53 | this.N = original.N;
|
---|
| 54 | this.R = original.R;
|
---|
| 55 | this.M = original.M;
|
---|
| 56 | }
|
---|
[11315] | 57 |
|
---|
[11901] | 58 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 59 | return new RFParameter(this, cloner);
|
---|
[11315] | 60 | }
|
---|
| 61 |
|
---|
[11901] | 62 | private IFixedValueParameter<IntValue> NParameter {
|
---|
| 63 | get { return (IFixedValueParameter<IntValue>)base["N"]; }
|
---|
| 64 | }
|
---|
[11315] | 65 |
|
---|
[11901] | 66 | private IFixedValueParameter<DoubleValue> RParameter {
|
---|
| 67 | get { return (IFixedValueParameter<DoubleValue>)base["R"]; }
|
---|
| 68 | }
|
---|
[11315] | 69 |
|
---|
[11901] | 70 | private IFixedValueParameter<DoubleValue> MParameter {
|
---|
| 71 | get { return (IFixedValueParameter<DoubleValue>)base["M"]; }
|
---|
[11315] | 72 | }
|
---|
| 73 |
|
---|
[11901] | 74 | public int N {
|
---|
| 75 | get { return NParameter.Value.Value; }
|
---|
| 76 | set { NParameter.Value.Value = value; }
|
---|
[11315] | 77 | }
|
---|
| 78 |
|
---|
[11901] | 79 | public double R {
|
---|
| 80 | get { return RParameter.Value.Value; }
|
---|
| 81 | set { RParameter.Value.Value = value; }
|
---|
| 82 | }
|
---|
[11315] | 83 |
|
---|
[11901] | 84 | public double M {
|
---|
| 85 | get { return MParameter.Value.Value; }
|
---|
| 86 | set { MParameter.Value.Value = value; }
|
---|
| 87 | }
|
---|
| 88 | }
|
---|
[11315] | 89 |
|
---|
[11901] | 90 | public static class RandomForestUtil {
|
---|
[17157] | 91 | public static void AssertParameters(double r, double m) {
|
---|
| 92 | if (r <= 0 || r > 1) throw new ArgumentException("The R parameter for random forest modeling must be between 0 and 1.");
|
---|
| 93 | if (m <= 0 || m > 1) throw new ArgumentException("The M parameter for random forest modeling must be between 0 and 1.");
|
---|
| 94 | }
|
---|
| 95 |
|
---|
| 96 | public static void AssertInputMatrix(double[,] inputMatrix) {
|
---|
| 97 | if (inputMatrix.ContainsNanOrInfinity())
|
---|
| 98 | throw new NotSupportedException("Random forest modeling does not support NaN or infinity values in the input dataset.");
|
---|
| 99 | }
|
---|
| 100 |
|
---|
| 101 | internal static alglib.decisionforest CreateRandomForestModel(int seed, double[,] inputMatrix, int nTrees, double r, double m, int nClasses, out alglib.dfreport rep) {
|
---|
| 102 | RandomForestUtil.AssertParameters(r, m);
|
---|
| 103 | RandomForestUtil.AssertInputMatrix(inputMatrix);
|
---|
| 104 |
|
---|
| 105 | int info = 0;
|
---|
| 106 | alglib.math.rndobject = new System.Random(seed);
|
---|
| 107 | var dForest = new alglib.decisionforest();
|
---|
| 108 | rep = new alglib.dfreport();
|
---|
| 109 | int nRows = inputMatrix.GetLength(0);
|
---|
| 110 | int nColumns = inputMatrix.GetLength(1);
|
---|
| 111 | int sampleSize = Math.Max((int)Math.Round(r * nRows), 1);
|
---|
| 112 | int nFeatures = Math.Max((int)Math.Round(m * (nColumns - 1)), 1);
|
---|
| 113 |
|
---|
| 114 | alglib.dforest.dfbuildinternal(inputMatrix, nRows, nColumns - 1, nClasses, nTrees, sampleSize, nFeatures, alglib.dforest.dfusestrongsplits + alglib.dforest.dfuseevs, ref info, dForest.innerobj, rep.innerobj);
|
---|
| 115 | if (info != 1) throw new ArgumentException("Error in calculation of random forest model");
|
---|
| 116 | return dForest;
|
---|
| 117 | }
|
---|
| 118 |
|
---|
| 119 |
|
---|
[11901] | 120 | private static void CrossValidate(IRegressionProblemData problemData, Tuple<IEnumerable<int>, IEnumerable<int>>[] partitions, int nTrees, double r, double m, int seed, out double avgTestMse) {
|
---|
| 121 | avgTestMse = 0;
|
---|
| 122 | var ds = problemData.Dataset;
|
---|
| 123 | var targetVariable = GetTargetVariableName(problemData);
|
---|
| 124 | foreach (var tuple in partitions) {
|
---|
[11315] | 125 | double rmsError, avgRelError, outOfBagAvgRelError, outOfBagRmsError;
|
---|
[11901] | 126 | var trainingRandomForestPartition = tuple.Item1;
|
---|
| 127 | var testRandomForestPartition = tuple.Item2;
|
---|
| 128 | var model = RandomForestModel.CreateRegressionModel(problemData, trainingRandomForestPartition, nTrees, r, m, seed, out rmsError, out avgRelError, out outOfBagRmsError, out outOfBagAvgRelError);
|
---|
| 129 | var estimatedValues = model.GetEstimatedValues(ds, testRandomForestPartition);
|
---|
| 130 | var targetValues = ds.GetDoubleValues(targetVariable, testRandomForestPartition);
|
---|
[11315] | 131 | OnlineCalculatorError calculatorError;
|
---|
[11901] | 132 | double mse = OnlineMeanSquaredErrorCalculator.Calculate(estimatedValues, targetValues, out calculatorError);
|
---|
[11315] | 133 | if (calculatorError != OnlineCalculatorError.None)
|
---|
| 134 | mse = double.NaN;
|
---|
[11901] | 135 | avgTestMse += mse;
|
---|
[11315] | 136 | }
|
---|
[11901] | 137 | avgTestMse /= partitions.Length;
|
---|
| 138 | }
|
---|
[11315] | 139 |
|
---|
[11901] | 140 | private static void CrossValidate(IClassificationProblemData problemData, Tuple<IEnumerable<int>, IEnumerable<int>>[] partitions, int nTrees, double r, double m, int seed, out double avgTestAccuracy) {
|
---|
| 141 | avgTestAccuracy = 0;
|
---|
| 142 | var ds = problemData.Dataset;
|
---|
| 143 | var targetVariable = GetTargetVariableName(problemData);
|
---|
| 144 | foreach (var tuple in partitions) {
|
---|
| 145 | double rmsError, avgRelError, outOfBagAvgRelError, outOfBagRmsError;
|
---|
| 146 | var trainingRandomForestPartition = tuple.Item1;
|
---|
| 147 | var testRandomForestPartition = tuple.Item2;
|
---|
| 148 | var model = RandomForestModel.CreateClassificationModel(problemData, trainingRandomForestPartition, nTrees, r, m, seed, out rmsError, out avgRelError, out outOfBagRmsError, out outOfBagAvgRelError);
|
---|
| 149 | var estimatedValues = model.GetEstimatedClassValues(ds, testRandomForestPartition);
|
---|
| 150 | var targetValues = ds.GetDoubleValues(targetVariable, testRandomForestPartition);
|
---|
| 151 | OnlineCalculatorError calculatorError;
|
---|
| 152 | double accuracy = OnlineAccuracyCalculator.Calculate(estimatedValues, targetValues, out calculatorError);
|
---|
| 153 | if (calculatorError != OnlineCalculatorError.None)
|
---|
| 154 | accuracy = double.NaN;
|
---|
| 155 | avgTestAccuracy += accuracy;
|
---|
| 156 | }
|
---|
| 157 | avgTestAccuracy /= partitions.Length;
|
---|
[11315] | 158 | }
|
---|
| 159 |
|
---|
[12702] | 160 | /// <summary>
|
---|
| 161 | /// Grid search without crossvalidation (since for random forests the out-of-bag estimate is unbiased)
|
---|
| 162 | /// </summary>
|
---|
| 163 | /// <param name="problemData">The regression problem data</param>
|
---|
| 164 | /// <param name="parameterRanges">The ranges for each parameter in the grid search</param>
|
---|
| 165 | /// <param name="seed">The random seed (required by the random forest model)</param>
|
---|
| 166 | /// <param name="maxDegreeOfParallelism">The maximum allowed number of threads (to parallelize the grid search)</param>
|
---|
[11901] | 167 | public static RFParameter GridSearch(IRegressionProblemData problemData, Dictionary<string, IEnumerable<double>> parameterRanges, int seed = 12345, int maxDegreeOfParallelism = 1) {
|
---|
| 168 | var setters = parameterRanges.Keys.Select(GenerateSetter).ToList();
|
---|
| 169 | var crossProduct = parameterRanges.Values.CartesianProduct();
|
---|
| 170 | double bestOutOfBagRmsError = double.MaxValue;
|
---|
| 171 | RFParameter bestParameters = new RFParameter();
|
---|
| 172 |
|
---|
[12702] | 173 | var locker = new object();
|
---|
[11901] | 174 | Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism }, parameterCombination => {
|
---|
| 175 | var parameterValues = parameterCombination.ToList();
|
---|
| 176 | var parameters = new RFParameter();
|
---|
| 177 | for (int i = 0; i < setters.Count; ++i) { setters[i](parameters, parameterValues[i]); }
|
---|
| 178 | double rmsError, outOfBagRmsError, avgRelError, outOfBagAvgRelError;
|
---|
| 179 | RandomForestModel.CreateRegressionModel(problemData, problemData.TrainingIndices, parameters.N, parameters.R, parameters.M, seed, out rmsError, out outOfBagRmsError, out avgRelError, out outOfBagAvgRelError);
|
---|
| 180 |
|
---|
| 181 | lock (locker) {
|
---|
| 182 | if (bestOutOfBagRmsError > outOfBagRmsError) {
|
---|
| 183 | bestOutOfBagRmsError = outOfBagRmsError;
|
---|
| 184 | bestParameters = (RFParameter)parameters.Clone();
|
---|
| 185 | }
|
---|
| 186 | }
|
---|
| 187 | });
|
---|
| 188 | return bestParameters;
|
---|
| 189 | }
|
---|
| 190 |
|
---|
[12702] | 191 | /// <summary>
|
---|
| 192 | /// Grid search without crossvalidation (since for random forests the out-of-bag estimate is unbiased)
|
---|
| 193 | /// </summary>
|
---|
| 194 | /// <param name="problemData">The classification problem data</param>
|
---|
| 195 | /// <param name="parameterRanges">The ranges for each parameter in the grid search</param>
|
---|
| 196 | /// <param name="seed">The random seed (required by the random forest model)</param>
|
---|
| 197 | /// <param name="maxDegreeOfParallelism">The maximum allowed number of threads (to parallelize the grid search)</param>
|
---|
[11901] | 198 | public static RFParameter GridSearch(IClassificationProblemData problemData, Dictionary<string, IEnumerable<double>> parameterRanges, int seed = 12345, int maxDegreeOfParallelism = 1) {
|
---|
| 199 | var setters = parameterRanges.Keys.Select(GenerateSetter).ToList();
|
---|
| 200 | var crossProduct = parameterRanges.Values.CartesianProduct();
|
---|
| 201 |
|
---|
| 202 | double bestOutOfBagRmsError = double.MaxValue;
|
---|
| 203 | RFParameter bestParameters = new RFParameter();
|
---|
| 204 |
|
---|
[12702] | 205 | var locker = new object();
|
---|
[11901] | 206 | Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism }, parameterCombination => {
|
---|
| 207 | var parameterValues = parameterCombination.ToList();
|
---|
| 208 | var parameters = new RFParameter();
|
---|
| 209 | for (int i = 0; i < setters.Count; ++i) { setters[i](parameters, parameterValues[i]); }
|
---|
| 210 | double rmsError, outOfBagRmsError, avgRelError, outOfBagAvgRelError;
|
---|
| 211 | RandomForestModel.CreateClassificationModel(problemData, problemData.TrainingIndices, parameters.N, parameters.R, parameters.M, seed,
|
---|
| 212 | out rmsError, out outOfBagRmsError, out avgRelError, out outOfBagAvgRelError);
|
---|
| 213 |
|
---|
| 214 | lock (locker) {
|
---|
| 215 | if (bestOutOfBagRmsError > outOfBagRmsError) {
|
---|
| 216 | bestOutOfBagRmsError = outOfBagRmsError;
|
---|
| 217 | bestParameters = (RFParameter)parameters.Clone();
|
---|
| 218 | }
|
---|
| 219 | }
|
---|
| 220 | });
|
---|
| 221 | return bestParameters;
|
---|
| 222 | }
|
---|
| 223 |
|
---|
[12702] | 224 | /// <summary>
|
---|
| 225 | /// Grid search with crossvalidation
|
---|
| 226 | /// </summary>
|
---|
| 227 | /// <param name="problemData">The regression problem data</param>
|
---|
| 228 | /// <param name="numberOfFolds">The number of folds for crossvalidation</param>
|
---|
| 229 | /// <param name="shuffleFolds">Specifies whether the folds should be shuffled</param>
|
---|
| 230 | /// <param name="parameterRanges">The ranges for each parameter in the grid search</param>
|
---|
| 231 | /// <param name="seed">The random seed (required by the random forest model)</param>
|
---|
| 232 | /// <param name="maxDegreeOfParallelism">The maximum allowed number of threads (to parallelize the grid search)</param>
|
---|
| 233 | /// <returns>The best parameter values found by the grid search</returns>
|
---|
[11901] | 234 | public static RFParameter GridSearch(IRegressionProblemData problemData, int numberOfFolds, bool shuffleFolds, Dictionary<string, IEnumerable<double>> parameterRanges, int seed = 12345, int maxDegreeOfParallelism = 1) {
|
---|
[11315] | 235 | DoubleValue mse = new DoubleValue(Double.MaxValue);
|
---|
[11901] | 236 | RFParameter bestParameter = new RFParameter();
|
---|
[11315] | 237 |
|
---|
[11901] | 238 | var setters = parameterRanges.Keys.Select(GenerateSetter).ToList();
|
---|
| 239 | var partitions = GenerateRandomForestPartitions(problemData, numberOfFolds);
|
---|
| 240 | var crossProduct = parameterRanges.Values.CartesianProduct();
|
---|
[11315] | 241 |
|
---|
[12702] | 242 | var locker = new object();
|
---|
[11901] | 243 | Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism }, parameterCombination => {
|
---|
| 244 | var parameterValues = parameterCombination.ToList();
|
---|
[11315] | 245 | double testMSE;
|
---|
| 246 | var parameters = new RFParameter();
|
---|
[11901] | 247 | for (int i = 0; i < setters.Count; ++i) {
|
---|
| 248 | setters[i](parameters, parameterValues[i]);
|
---|
[11315] | 249 | }
|
---|
[11901] | 250 | CrossValidate(problemData, partitions, parameters.N, parameters.R, parameters.M, seed, out testMSE);
|
---|
| 251 |
|
---|
| 252 | lock (locker) {
|
---|
| 253 | if (testMSE < mse.Value) {
|
---|
[11315] | 254 | mse.Value = testMSE;
|
---|
[11901] | 255 | bestParameter = (RFParameter)parameters.Clone();
|
---|
[11315] | 256 | }
|
---|
[11901] | 257 | }
|
---|
| 258 | });
|
---|
| 259 | return bestParameter;
|
---|
| 260 | }
|
---|
| 261 |
|
---|
[12702] | 262 | /// <summary>
|
---|
| 263 | /// Grid search with crossvalidation
|
---|
| 264 | /// </summary>
|
---|
| 265 | /// <param name="problemData">The classification problem data</param>
|
---|
| 266 | /// <param name="numberOfFolds">The number of folds for crossvalidation</param>
|
---|
| 267 | /// <param name="shuffleFolds">Specifies whether the folds should be shuffled</param>
|
---|
| 268 | /// <param name="parameterRanges">The ranges for each parameter in the grid search</param>
|
---|
| 269 | /// <param name="seed">The random seed (for shuffling)</param>
|
---|
| 270 | /// <param name="maxDegreeOfParallelism">The maximum allowed number of threads (to parallelize the grid search)</param>
|
---|
[11901] | 271 | public static RFParameter GridSearch(IClassificationProblemData problemData, int numberOfFolds, bool shuffleFolds, Dictionary<string, IEnumerable<double>> parameterRanges, int seed = 12345, int maxDegreeOfParallelism = 1) {
|
---|
| 272 | DoubleValue accuracy = new DoubleValue(0);
|
---|
| 273 | RFParameter bestParameter = new RFParameter();
|
---|
| 274 |
|
---|
| 275 | var setters = parameterRanges.Keys.Select(GenerateSetter).ToList();
|
---|
| 276 | var crossProduct = parameterRanges.Values.CartesianProduct();
|
---|
| 277 | var partitions = GenerateRandomForestPartitions(problemData, numberOfFolds, shuffleFolds);
|
---|
| 278 |
|
---|
[12702] | 279 | var locker = new object();
|
---|
[11901] | 280 | Parallel.ForEach(crossProduct, new ParallelOptions { MaxDegreeOfParallelism = maxDegreeOfParallelism }, parameterCombination => {
|
---|
| 281 | var parameterValues = parameterCombination.ToList();
|
---|
| 282 | double testAccuracy;
|
---|
| 283 | var parameters = new RFParameter();
|
---|
| 284 | for (int i = 0; i < setters.Count; ++i) {
|
---|
| 285 | setters[i](parameters, parameterValues[i]);
|
---|
| 286 | }
|
---|
| 287 | CrossValidate(problemData, partitions, parameters.N, parameters.R, parameters.M, seed, out testAccuracy);
|
---|
| 288 |
|
---|
| 289 | lock (locker) {
|
---|
| 290 | if (testAccuracy > accuracy.Value) {
|
---|
| 291 | accuracy.Value = testAccuracy;
|
---|
[11315] | 292 | bestParameter = (RFParameter)parameters.Clone();
|
---|
| 293 | }
|
---|
| 294 | }
|
---|
| 295 | });
|
---|
| 296 | return bestParameter;
|
---|
| 297 | }
|
---|
[11901] | 298 |
|
---|
| 299 | private static Tuple<IEnumerable<int>, IEnumerable<int>>[] GenerateRandomForestPartitions(IDataAnalysisProblemData problemData, int numberOfFolds, bool shuffleFolds = false) {
|
---|
| 300 | var folds = GenerateFolds(problemData, numberOfFolds, shuffleFolds).ToList();
|
---|
| 301 | var partitions = new Tuple<IEnumerable<int>, IEnumerable<int>>[numberOfFolds];
|
---|
| 302 |
|
---|
| 303 | for (int i = 0; i < numberOfFolds; ++i) {
|
---|
| 304 | int p = i; // avoid "access to modified closure" warning
|
---|
| 305 | var trainingRows = folds.SelectMany((par, j) => j != p ? par : Enumerable.Empty<int>());
|
---|
| 306 | var testRows = folds[i];
|
---|
| 307 | partitions[i] = new Tuple<IEnumerable<int>, IEnumerable<int>>(trainingRows, testRows);
|
---|
| 308 | }
|
---|
| 309 | return partitions;
|
---|
| 310 | }
|
---|
| 311 |
|
---|
| 312 | public static IEnumerable<IEnumerable<int>> GenerateFolds(IDataAnalysisProblemData problemData, int numberOfFolds, bool shuffleFolds = false) {
|
---|
| 313 | var random = new MersenneTwister((uint)Environment.TickCount);
|
---|
| 314 | if (problemData is IRegressionProblemData) {
|
---|
| 315 | var trainingIndices = shuffleFolds ? problemData.TrainingIndices.OrderBy(x => random.Next()) : problemData.TrainingIndices;
|
---|
| 316 | return GenerateFolds(trainingIndices, problemData.TrainingPartition.Size, numberOfFolds);
|
---|
| 317 | }
|
---|
| 318 | if (problemData is IClassificationProblemData) {
|
---|
| 319 | // when shuffle is enabled do stratified folds generation, some folds may have zero elements
|
---|
| 320 | // otherwise, generate folds normally
|
---|
| 321 | return shuffleFolds ? GenerateFoldsStratified(problemData as IClassificationProblemData, numberOfFolds, random) : GenerateFolds(problemData.TrainingIndices, problemData.TrainingPartition.Size, numberOfFolds);
|
---|
| 322 | }
|
---|
| 323 | throw new ArgumentException("Problem data is neither regression or classification problem data.");
|
---|
| 324 | }
|
---|
| 325 |
|
---|
| 326 | /// <summary>
|
---|
| 327 | /// Stratified fold generation from classification data. Stratification means that we ensure the same distribution of class labels for each fold.
|
---|
| 328 | /// The samples are grouped by class label and each group is split into @numberOfFolds parts. The final folds are formed from the joining of
|
---|
| 329 | /// the corresponding parts from each class label.
|
---|
| 330 | /// </summary>
|
---|
| 331 | /// <param name="problemData">The classification problem data.</param>
|
---|
| 332 | /// <param name="numberOfFolds">The number of folds in which to split the data.</param>
|
---|
| 333 | /// <param name="random">The random generator used to shuffle the folds.</param>
|
---|
| 334 | /// <returns>An enumerable sequece of folds, where a fold is represented by a sequence of row indices.</returns>
|
---|
| 335 | private static IEnumerable<IEnumerable<int>> GenerateFoldsStratified(IClassificationProblemData problemData, int numberOfFolds, IRandom random) {
|
---|
| 336 | var values = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, problemData.TrainingIndices);
|
---|
| 337 | var valuesIndices = problemData.TrainingIndices.Zip(values, (i, v) => new { Index = i, Value = v }).ToList();
|
---|
| 338 | IEnumerable<IEnumerable<IEnumerable<int>>> foldsByClass = valuesIndices.GroupBy(x => x.Value, x => x.Index).Select(g => GenerateFolds(g, g.Count(), numberOfFolds));
|
---|
| 339 | var enumerators = foldsByClass.Select(f => f.GetEnumerator()).ToList();
|
---|
| 340 | while (enumerators.All(e => e.MoveNext())) {
|
---|
| 341 | yield return enumerators.SelectMany(e => e.Current).OrderBy(x => random.Next()).ToList();
|
---|
| 342 | }
|
---|
| 343 | }
|
---|
| 344 |
|
---|
| 345 | private static IEnumerable<IEnumerable<T>> GenerateFolds<T>(IEnumerable<T> values, int valuesCount, int numberOfFolds) {
|
---|
| 346 | // if number of folds is greater than the number of values, some empty folds will be returned
|
---|
| 347 | if (valuesCount < numberOfFolds) {
|
---|
| 348 | for (int i = 0; i < numberOfFolds; ++i)
|
---|
| 349 | yield return i < valuesCount ? values.Skip(i).Take(1) : Enumerable.Empty<T>();
|
---|
| 350 | } else {
|
---|
| 351 | int f = valuesCount / numberOfFolds, r = valuesCount % numberOfFolds; // number of folds rounded to integer and remainder
|
---|
| 352 | int start = 0, end = f;
|
---|
| 353 | for (int i = 0; i < numberOfFolds; ++i) {
|
---|
| 354 | if (r > 0) {
|
---|
| 355 | ++end;
|
---|
| 356 | --r;
|
---|
| 357 | }
|
---|
| 358 | yield return values.Skip(start).Take(end - start);
|
---|
| 359 | start = end;
|
---|
| 360 | end += f;
|
---|
| 361 | }
|
---|
| 362 | }
|
---|
| 363 | }
|
---|
| 364 |
|
---|
| 365 | private static Action<RFParameter, double> GenerateSetter(string field) {
|
---|
| 366 | var targetExp = Expression.Parameter(typeof(RFParameter));
|
---|
| 367 | var valueExp = Expression.Parameter(typeof(double));
|
---|
| 368 | var fieldExp = Expression.Property(targetExp, field);
|
---|
| 369 | var assignExp = Expression.Assign(fieldExp, Expression.Convert(valueExp, fieldExp.Type));
|
---|
| 370 | var setter = Expression.Lambda<Action<RFParameter, double>>(assignExp, targetExp, valueExp).Compile();
|
---|
| 371 | return setter;
|
---|
| 372 | }
|
---|
| 373 |
|
---|
| 374 | private static string GetTargetVariableName(IDataAnalysisProblemData problemData) {
|
---|
| 375 | var regressionProblemData = problemData as IRegressionProblemData;
|
---|
| 376 | var classificationProblemData = problemData as IClassificationProblemData;
|
---|
| 377 |
|
---|
| 378 | if (regressionProblemData != null)
|
---|
| 379 | return regressionProblemData.TargetVariable;
|
---|
| 380 | if (classificationProblemData != null)
|
---|
| 381 | return classificationProblemData.TargetVariable;
|
---|
| 382 |
|
---|
| 383 | throw new ArgumentException("Problem data is neither regression or classification problem data.");
|
---|
| 384 | }
|
---|
[11315] | 385 | }
|
---|
| 386 | }
|
---|