[17154] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[17181] | 3 | * Copyright (C) Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[17154] | 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;
|
---|
| 23 | using System.Collections.Generic;
|
---|
| 24 | using System.Linq;
|
---|
| 25 | using HEAL.Attic;
|
---|
| 26 | using HeuristicLab.Common;
|
---|
| 27 | using HeuristicLab.Core;
|
---|
| 28 | using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
|
---|
| 29 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 30 | using HeuristicLab.Problems.DataAnalysis.Symbolic;
|
---|
| 31 |
|
---|
| 32 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
| 33 | [StorableType("9C797DF0-1169-4381-A732-6DAB90802839")]
|
---|
| 34 | [Item("RandomForestModelFull", "Represents a random forest for regression and classification.")]
|
---|
| 35 | public sealed class RandomForestModelFull : ClassificationModel, IRandomForestModel {
|
---|
| 36 |
|
---|
| 37 | public override IEnumerable<string> VariablesUsedForPrediction {
|
---|
| 38 | get { return inputVariables; }
|
---|
| 39 | }
|
---|
| 40 |
|
---|
| 41 | [Storable]
|
---|
| 42 | private double[] classValues;
|
---|
| 43 |
|
---|
| 44 | [Storable]
|
---|
| 45 | private string[] inputVariables;
|
---|
| 46 |
|
---|
| 47 | public int NumberOfTrees {
|
---|
| 48 | get { return RandomForestNTrees; }
|
---|
| 49 | }
|
---|
| 50 |
|
---|
| 51 | // not persisted
|
---|
| 52 | private alglib.decisionforest randomForest;
|
---|
| 53 |
|
---|
| 54 | [Storable]
|
---|
| 55 | private int RandomForestBufSize {
|
---|
| 56 | get { return randomForest.innerobj.bufsize; }
|
---|
| 57 | set { randomForest.innerobj.bufsize = value; }
|
---|
| 58 | }
|
---|
| 59 | [Storable]
|
---|
| 60 | private int RandomForestNClasses {
|
---|
| 61 | get { return randomForest.innerobj.nclasses; }
|
---|
| 62 | set { randomForest.innerobj.nclasses = value; }
|
---|
| 63 | }
|
---|
| 64 | [Storable]
|
---|
| 65 | private int RandomForestNTrees {
|
---|
| 66 | get { return randomForest.innerobj.ntrees; }
|
---|
| 67 | set { randomForest.innerobj.ntrees = value; }
|
---|
| 68 | }
|
---|
| 69 | [Storable]
|
---|
| 70 | private int RandomForestNVars {
|
---|
| 71 | get { return randomForest.innerobj.nvars; }
|
---|
| 72 | set { randomForest.innerobj.nvars = value; }
|
---|
| 73 | }
|
---|
| 74 | [Storable]
|
---|
| 75 | private double[] RandomForestTrees {
|
---|
| 76 | get { return randomForest.innerobj.trees; }
|
---|
| 77 | set { randomForest.innerobj.trees = value; }
|
---|
| 78 | }
|
---|
| 79 |
|
---|
| 80 | [StorableConstructor]
|
---|
| 81 | private RandomForestModelFull(StorableConstructorFlag _) : base(_) {
|
---|
| 82 | randomForest = new alglib.decisionforest();
|
---|
| 83 | }
|
---|
| 84 |
|
---|
| 85 | private RandomForestModelFull(RandomForestModelFull original, Cloner cloner) : base(original, cloner) {
|
---|
| 86 | randomForest = new alglib.decisionforest();
|
---|
| 87 | randomForest.innerobj.bufsize = original.randomForest.innerobj.bufsize;
|
---|
| 88 | randomForest.innerobj.nclasses = original.randomForest.innerobj.nclasses;
|
---|
| 89 | randomForest.innerobj.ntrees = original.randomForest.innerobj.ntrees;
|
---|
| 90 | randomForest.innerobj.nvars = original.randomForest.innerobj.nvars;
|
---|
| 91 | randomForest.innerobj.trees = (double[])original.randomForest.innerobj.trees.Clone();
|
---|
| 92 |
|
---|
| 93 | // following fields are immutable so we don't need to clone them
|
---|
| 94 | inputVariables = original.inputVariables;
|
---|
| 95 | classValues = original.classValues;
|
---|
| 96 | }
|
---|
| 97 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 98 | return new RandomForestModelFull(this, cloner);
|
---|
| 99 | }
|
---|
| 100 |
|
---|
| 101 | public RandomForestModelFull(alglib.decisionforest decisionForest, string targetVariable, IEnumerable<string> inputVariables, IEnumerable<double> classValues = null) : base(targetVariable) {
|
---|
| 102 | this.name = ItemName;
|
---|
| 103 | this.description = ItemDescription;
|
---|
| 104 |
|
---|
| 105 | randomForest = decisionForest;
|
---|
| 106 |
|
---|
| 107 | this.inputVariables = inputVariables.ToArray();
|
---|
| 108 |
|
---|
| 109 | //classValues are only use for classification models
|
---|
| 110 | if (classValues == null) this.classValues = new double[0];
|
---|
| 111 | else this.classValues = classValues.ToArray();
|
---|
| 112 | }
|
---|
| 113 |
|
---|
| 114 |
|
---|
| 115 | public IRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
| 116 | return new RandomForestRegressionSolution(this, new RegressionProblemData(problemData));
|
---|
| 117 | }
|
---|
| 118 | public override IClassificationSolution CreateClassificationSolution(IClassificationProblemData problemData) {
|
---|
| 119 | return new RandomForestClassificationSolution(this, new ClassificationProblemData(problemData));
|
---|
| 120 | }
|
---|
| 121 |
|
---|
| 122 | public bool IsProblemDataCompatible(IRegressionProblemData problemData, out string errorMessage) {
|
---|
| 123 | return RegressionModel.IsProblemDataCompatible(this, problemData, out errorMessage);
|
---|
| 124 | }
|
---|
| 125 |
|
---|
| 126 | public override bool IsProblemDataCompatible(IDataAnalysisProblemData problemData, out string errorMessage) {
|
---|
| 127 | if (problemData == null) throw new ArgumentNullException("problemData", "The provided problemData is null.");
|
---|
| 128 |
|
---|
| 129 | var regressionProblemData = problemData as IRegressionProblemData;
|
---|
| 130 | if (regressionProblemData != null)
|
---|
| 131 | return IsProblemDataCompatible(regressionProblemData, out errorMessage);
|
---|
| 132 |
|
---|
| 133 | var classificationProblemData = problemData as IClassificationProblemData;
|
---|
| 134 | if (classificationProblemData != null)
|
---|
| 135 | return IsProblemDataCompatible(classificationProblemData, out errorMessage);
|
---|
| 136 |
|
---|
| 137 | throw new ArgumentException("The problem data is not compatible with this random forest. Instead a " + problemData.GetType().GetPrettyName() + " was provided.", "problemData");
|
---|
| 138 | }
|
---|
| 139 |
|
---|
| 140 | public IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
| 141 | double[,] inputData = dataset.ToArray(inputVariables, rows);
|
---|
| 142 | RandomForestUtil.AssertInputMatrix(inputData);
|
---|
| 143 |
|
---|
| 144 | int n = inputData.GetLength(0);
|
---|
| 145 | int columns = inputData.GetLength(1);
|
---|
| 146 | double[] x = new double[columns];
|
---|
| 147 | double[] y = new double[1];
|
---|
| 148 |
|
---|
| 149 | for (int row = 0; row < n; row++) {
|
---|
| 150 | for (int column = 0; column < columns; column++) {
|
---|
| 151 | x[column] = inputData[row, column];
|
---|
| 152 | }
|
---|
| 153 | alglib.dfprocess(randomForest, x, ref y);
|
---|
| 154 | yield return y[0];
|
---|
| 155 | }
|
---|
| 156 | }
|
---|
| 157 | public IEnumerable<double> GetEstimatedVariances(IDataset dataset, IEnumerable<int> rows) {
|
---|
| 158 | double[,] inputData = dataset.ToArray(inputVariables, rows);
|
---|
| 159 | RandomForestUtil.AssertInputMatrix(inputData);
|
---|
| 160 |
|
---|
| 161 | int n = inputData.GetLength(0);
|
---|
| 162 | int columns = inputData.GetLength(1);
|
---|
| 163 | double[] x = new double[columns];
|
---|
| 164 | double[] ys = new double[this.randomForest.innerobj.ntrees];
|
---|
| 165 |
|
---|
| 166 | for (int row = 0; row < n; row++) {
|
---|
| 167 | for (int column = 0; column < columns; column++) {
|
---|
| 168 | x[column] = inputData[row, column];
|
---|
| 169 | }
|
---|
| 170 | alglib.dforest.dfprocessraw(randomForest.innerobj, x, ref ys);
|
---|
| 171 | yield return ys.VariancePop();
|
---|
| 172 | }
|
---|
| 173 | }
|
---|
| 174 |
|
---|
| 175 | public override IEnumerable<double> GetEstimatedClassValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
| 176 | double[,] inputData = dataset.ToArray(inputVariables, rows);
|
---|
| 177 | RandomForestUtil.AssertInputMatrix(inputData);
|
---|
| 178 |
|
---|
| 179 | int n = inputData.GetLength(0);
|
---|
| 180 | int columns = inputData.GetLength(1);
|
---|
| 181 | double[] x = new double[columns];
|
---|
| 182 | double[] y = new double[randomForest.innerobj.nclasses];
|
---|
| 183 |
|
---|
| 184 | for (int row = 0; row < n; row++) {
|
---|
| 185 | for (int column = 0; column < columns; column++) {
|
---|
| 186 | x[column] = inputData[row, column];
|
---|
| 187 | }
|
---|
| 188 | alglib.dfprocess(randomForest, x, ref y);
|
---|
| 189 | // find class for with the largest probability value
|
---|
| 190 | int maxProbClassIndex = 0;
|
---|
| 191 | double maxProb = y[0];
|
---|
| 192 | for (int i = 1; i < y.Length; i++) {
|
---|
| 193 | if (maxProb < y[i]) {
|
---|
| 194 | maxProb = y[i];
|
---|
| 195 | maxProbClassIndex = i;
|
---|
| 196 | }
|
---|
| 197 | }
|
---|
| 198 | yield return classValues[maxProbClassIndex];
|
---|
| 199 | }
|
---|
| 200 | }
|
---|
| 201 |
|
---|
| 202 | public ISymbolicExpressionTree ExtractTree(int treeIdx) {
|
---|
| 203 | var rf = randomForest;
|
---|
| 204 | // hoping that the internal representation of alglib is stable
|
---|
| 205 |
|
---|
| 206 | // TREE FORMAT
|
---|
| 207 | // W[Offs] - size of sub-array (for the tree)
|
---|
| 208 | // node info:
|
---|
| 209 | // W[K+0] - variable number (-1 for leaf mode)
|
---|
| 210 | // W[K+1] - threshold (class/value for leaf node)
|
---|
| 211 | // W[K+2] - ">=" branch index (absent for leaf node)
|
---|
| 212 |
|
---|
| 213 | // skip irrelevant trees
|
---|
| 214 | int offset = 0;
|
---|
| 215 | for (int i = 0; i < treeIdx - 1; i++) {
|
---|
| 216 | offset = offset + (int)Math.Round(rf.innerobj.trees[offset]);
|
---|
| 217 | }
|
---|
| 218 |
|
---|
| 219 | var constSy = new Constant();
|
---|
| 220 | var varCondSy = new VariableCondition() { IgnoreSlope = true };
|
---|
| 221 |
|
---|
| 222 | var node = CreateRegressionTreeRec(rf.innerobj.trees, offset, offset + 1, constSy, varCondSy);
|
---|
| 223 |
|
---|
| 224 | var startNode = new StartSymbol().CreateTreeNode();
|
---|
| 225 | startNode.AddSubtree(node);
|
---|
| 226 | var root = new ProgramRootSymbol().CreateTreeNode();
|
---|
| 227 | root.AddSubtree(startNode);
|
---|
| 228 | return new SymbolicExpressionTree(root);
|
---|
| 229 | }
|
---|
| 230 |
|
---|
| 231 | private ISymbolicExpressionTreeNode CreateRegressionTreeRec(double[] trees, int offset, int k, Constant constSy, VariableCondition varCondSy) {
|
---|
| 232 |
|
---|
| 233 | // alglib source for evaluation of one tree (dfprocessinternal)
|
---|
| 234 | // offs = 0
|
---|
| 235 | //
|
---|
| 236 | // Set pointer to the root
|
---|
| 237 | //
|
---|
| 238 | // k = offs + 1;
|
---|
| 239 | //
|
---|
| 240 | // //
|
---|
| 241 | // // Navigate through the tree
|
---|
| 242 | // //
|
---|
| 243 | // while (true) {
|
---|
| 244 | // if ((double)(df.trees[k]) == (double)(-1)) {
|
---|
| 245 | // if (df.nclasses == 1) {
|
---|
| 246 | // y[0] = y[0] + df.trees[k + 1];
|
---|
| 247 | // } else {
|
---|
| 248 | // idx = (int)Math.Round(df.trees[k + 1]);
|
---|
| 249 | // y[idx] = y[idx] + 1;
|
---|
| 250 | // }
|
---|
| 251 | // break;
|
---|
| 252 | // }
|
---|
| 253 | // if ((double)(x[(int)Math.Round(df.trees[k])]) < (double)(df.trees[k + 1])) {
|
---|
| 254 | // k = k + innernodewidth;
|
---|
| 255 | // } else {
|
---|
| 256 | // k = offs + (int)Math.Round(df.trees[k + 2]);
|
---|
| 257 | // }
|
---|
| 258 | // }
|
---|
| 259 |
|
---|
| 260 | if ((double)(trees[k]) == (double)(-1)) {
|
---|
| 261 | var constNode = (ConstantTreeNode)constSy.CreateTreeNode();
|
---|
| 262 | constNode.Value = trees[k + 1];
|
---|
| 263 | return constNode;
|
---|
| 264 | } else {
|
---|
| 265 | var condNode = (VariableConditionTreeNode)varCondSy.CreateTreeNode();
|
---|
| 266 | condNode.VariableName = inputVariables[(int)Math.Round(trees[k])];
|
---|
| 267 | condNode.Threshold = trees[k + 1];
|
---|
| 268 | condNode.Slope = double.PositiveInfinity;
|
---|
| 269 |
|
---|
| 270 | var left = CreateRegressionTreeRec(trees, offset, k + 3, constSy, varCondSy);
|
---|
| 271 | var right = CreateRegressionTreeRec(trees, offset, offset + (int)Math.Round(trees[k + 2]), constSy, varCondSy);
|
---|
| 272 |
|
---|
| 273 | condNode.AddSubtree(left); // not 100% correct because interpreter uses: if(x <= thres) left() else right() and RF uses if(x < thres) left() else right() (see above)
|
---|
| 274 | condNode.AddSubtree(right);
|
---|
| 275 | return condNode;
|
---|
| 276 | }
|
---|
| 277 | }
|
---|
| 278 | }
|
---|
| 279 | } |
---|