[6583] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[12012] | 3 | * Copyright (C) 2002-2015 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[6583] | 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 HeuristicLab.Common;
|
---|
| 26 | using HeuristicLab.Core;
|
---|
| 27 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 28 | using HeuristicLab.Problems.DataAnalysis;
|
---|
| 29 |
|
---|
| 30 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
| 31 | /// <summary>
|
---|
| 32 | /// Represents a nearest neighbour model for regression and classification
|
---|
| 33 | /// </summary>
|
---|
[13368] | 34 | [StorableClass("4AA8CBF7-4CE7-487F-8E35-67AA1A104E8F")]
|
---|
[8465] | 35 | [Item("NearestNeighbourModel", "Represents a nearest neighbour model for regression and classification.")]
|
---|
[6583] | 36 | public sealed class NearestNeighbourModel : NamedItem, INearestNeighbourModel {
|
---|
| 37 |
|
---|
| 38 | private alglib.nearestneighbor.kdtree kdTree;
|
---|
| 39 | public alglib.nearestneighbor.kdtree KDTree {
|
---|
| 40 | get { return kdTree; }
|
---|
| 41 | set {
|
---|
| 42 | if (value != kdTree) {
|
---|
| 43 | if (value == null) throw new ArgumentNullException();
|
---|
| 44 | kdTree = value;
|
---|
| 45 | OnChanged(EventArgs.Empty);
|
---|
| 46 | }
|
---|
| 47 | }
|
---|
| 48 | }
|
---|
| 49 |
|
---|
| 50 | [Storable]
|
---|
| 51 | private string targetVariable;
|
---|
| 52 | [Storable]
|
---|
| 53 | private string[] allowedInputVariables;
|
---|
| 54 | [Storable]
|
---|
| 55 | private double[] classValues;
|
---|
| 56 | [Storable]
|
---|
| 57 | private int k;
|
---|
[8465] | 58 |
|
---|
[6583] | 59 | [StorableConstructor]
|
---|
| 60 | private NearestNeighbourModel(bool deserializing)
|
---|
| 61 | : base(deserializing) {
|
---|
| 62 | if (deserializing)
|
---|
| 63 | kdTree = new alglib.nearestneighbor.kdtree();
|
---|
| 64 | }
|
---|
| 65 | private NearestNeighbourModel(NearestNeighbourModel original, Cloner cloner)
|
---|
| 66 | : base(original, cloner) {
|
---|
| 67 | kdTree = new alglib.nearestneighbor.kdtree();
|
---|
| 68 | kdTree.approxf = original.kdTree.approxf;
|
---|
| 69 | kdTree.boxmax = (double[])original.kdTree.boxmax.Clone();
|
---|
| 70 | kdTree.boxmin = (double[])original.kdTree.boxmin.Clone();
|
---|
| 71 | kdTree.buf = (double[])original.kdTree.buf.Clone();
|
---|
| 72 | kdTree.curboxmax = (double[])original.kdTree.curboxmax.Clone();
|
---|
| 73 | kdTree.curboxmin = (double[])original.kdTree.curboxmin.Clone();
|
---|
| 74 | kdTree.curdist = original.kdTree.curdist;
|
---|
| 75 | kdTree.debugcounter = original.kdTree.debugcounter;
|
---|
| 76 | kdTree.idx = (int[])original.kdTree.idx.Clone();
|
---|
| 77 | kdTree.kcur = original.kdTree.kcur;
|
---|
| 78 | kdTree.kneeded = original.kdTree.kneeded;
|
---|
| 79 | kdTree.n = original.kdTree.n;
|
---|
| 80 | kdTree.nodes = (int[])original.kdTree.nodes.Clone();
|
---|
| 81 | kdTree.normtype = original.kdTree.normtype;
|
---|
| 82 | kdTree.nx = original.kdTree.nx;
|
---|
| 83 | kdTree.ny = original.kdTree.ny;
|
---|
| 84 | kdTree.r = (double[])original.kdTree.r.Clone();
|
---|
| 85 | kdTree.rneeded = original.kdTree.rneeded;
|
---|
| 86 | kdTree.selfmatch = original.kdTree.selfmatch;
|
---|
| 87 | kdTree.splits = (double[])original.kdTree.splits.Clone();
|
---|
| 88 | kdTree.tags = (int[])original.kdTree.tags.Clone();
|
---|
| 89 | kdTree.x = (double[])original.kdTree.x.Clone();
|
---|
| 90 | kdTree.xy = (double[,])original.kdTree.xy.Clone();
|
---|
| 91 |
|
---|
| 92 | k = original.k;
|
---|
| 93 | targetVariable = original.targetVariable;
|
---|
| 94 | allowedInputVariables = (string[])original.allowedInputVariables.Clone();
|
---|
| 95 | if (original.classValues != null)
|
---|
| 96 | this.classValues = (double[])original.classValues.Clone();
|
---|
| 97 | }
|
---|
[12509] | 98 | public NearestNeighbourModel(IDataset dataset, IEnumerable<int> rows, int k, string targetVariable, IEnumerable<string> allowedInputVariables, double[] classValues = null) {
|
---|
[8467] | 99 | Name = ItemName;
|
---|
| 100 | Description = ItemDescription;
|
---|
[6583] | 101 | this.k = k;
|
---|
| 102 | this.targetVariable = targetVariable;
|
---|
| 103 | this.allowedInputVariables = allowedInputVariables.ToArray();
|
---|
[8465] | 104 |
|
---|
| 105 | var inputMatrix = AlglibUtil.PrepareInputMatrix(dataset,
|
---|
| 106 | allowedInputVariables.Concat(new string[] { targetVariable }),
|
---|
| 107 | rows);
|
---|
| 108 |
|
---|
| 109 | if (inputMatrix.Cast<double>().Any(x => double.IsNaN(x) || double.IsInfinity(x)))
|
---|
| 110 | throw new NotSupportedException(
|
---|
| 111 | "Nearest neighbour classification does not support NaN or infinity values in the input dataset.");
|
---|
| 112 |
|
---|
| 113 | this.kdTree = new alglib.nearestneighbor.kdtree();
|
---|
| 114 |
|
---|
| 115 | var nRows = inputMatrix.GetLength(0);
|
---|
| 116 | var nFeatures = inputMatrix.GetLength(1) - 1;
|
---|
| 117 |
|
---|
| 118 | if (classValues != null) {
|
---|
[6583] | 119 | this.classValues = (double[])classValues.Clone();
|
---|
[8465] | 120 | int nClasses = classValues.Length;
|
---|
| 121 | // map original class values to values [0..nClasses-1]
|
---|
| 122 | var classIndices = new Dictionary<double, double>();
|
---|
| 123 | for (int i = 0; i < nClasses; i++)
|
---|
| 124 | classIndices[classValues[i]] = i;
|
---|
| 125 |
|
---|
| 126 | for (int row = 0; row < nRows; row++) {
|
---|
| 127 | inputMatrix[row, nFeatures] = classIndices[inputMatrix[row, nFeatures]];
|
---|
| 128 | }
|
---|
| 129 | }
|
---|
| 130 | alglib.nearestneighbor.kdtreebuild(inputMatrix, nRows, inputMatrix.GetLength(1) - 1, 1, 2, kdTree);
|
---|
[6583] | 131 | }
|
---|
| 132 |
|
---|
| 133 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 134 | return new NearestNeighbourModel(this, cloner);
|
---|
| 135 | }
|
---|
| 136 |
|
---|
[12509] | 137 | public IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
[6583] | 138 | double[,] inputData = AlglibUtil.PrepareInputMatrix(dataset, allowedInputVariables, rows);
|
---|
| 139 |
|
---|
| 140 | int n = inputData.GetLength(0);
|
---|
| 141 | int columns = inputData.GetLength(1);
|
---|
| 142 | double[] x = new double[columns];
|
---|
| 143 | double[] y = new double[1];
|
---|
| 144 | double[] dists = new double[k];
|
---|
| 145 | double[,] neighbours = new double[k, columns + 1];
|
---|
| 146 |
|
---|
| 147 | for (int row = 0; row < n; row++) {
|
---|
| 148 | for (int column = 0; column < columns; column++) {
|
---|
| 149 | x[column] = inputData[row, column];
|
---|
| 150 | }
|
---|
| 151 | int actNeighbours = alglib.nearestneighbor.kdtreequeryknn(kdTree, x, k, false);
|
---|
| 152 | alglib.nearestneighbor.kdtreequeryresultsdistances(kdTree, ref dists);
|
---|
| 153 | alglib.nearestneighbor.kdtreequeryresultsxy(kdTree, ref neighbours);
|
---|
| 154 |
|
---|
| 155 | double distanceWeightedValue = 0.0;
|
---|
| 156 | double distsSum = 0.0;
|
---|
| 157 | for (int i = 0; i < actNeighbours; i++) {
|
---|
| 158 | distanceWeightedValue += neighbours[i, columns] / dists[i];
|
---|
| 159 | distsSum += 1.0 / dists[i];
|
---|
| 160 | }
|
---|
| 161 | yield return distanceWeightedValue / distsSum;
|
---|
| 162 | }
|
---|
| 163 | }
|
---|
| 164 |
|
---|
[12509] | 165 | public IEnumerable<double> GetEstimatedClassValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
[8465] | 166 | if (classValues == null) throw new InvalidOperationException("No class values are defined.");
|
---|
[6583] | 167 | double[,] inputData = AlglibUtil.PrepareInputMatrix(dataset, allowedInputVariables, rows);
|
---|
| 168 |
|
---|
| 169 | int n = inputData.GetLength(0);
|
---|
| 170 | int columns = inputData.GetLength(1);
|
---|
| 171 | double[] x = new double[columns];
|
---|
| 172 | int[] y = new int[classValues.Length];
|
---|
| 173 | double[] dists = new double[k];
|
---|
| 174 | double[,] neighbours = new double[k, columns + 1];
|
---|
| 175 |
|
---|
| 176 | for (int row = 0; row < n; row++) {
|
---|
| 177 | for (int column = 0; column < columns; column++) {
|
---|
| 178 | x[column] = inputData[row, column];
|
---|
| 179 | }
|
---|
| 180 | int actNeighbours = alglib.nearestneighbor.kdtreequeryknn(kdTree, x, k, false);
|
---|
| 181 | alglib.nearestneighbor.kdtreequeryresultsdistances(kdTree, ref dists);
|
---|
| 182 | alglib.nearestneighbor.kdtreequeryresultsxy(kdTree, ref neighbours);
|
---|
| 183 |
|
---|
| 184 | Array.Clear(y, 0, y.Length);
|
---|
| 185 | for (int i = 0; i < actNeighbours; i++) {
|
---|
| 186 | int classValue = (int)Math.Round(neighbours[i, columns]);
|
---|
| 187 | y[classValue]++;
|
---|
| 188 | }
|
---|
| 189 |
|
---|
| 190 | // find class for with the largest probability value
|
---|
| 191 | int maxProbClassIndex = 0;
|
---|
| 192 | double maxProb = y[0];
|
---|
| 193 | for (int i = 1; i < y.Length; i++) {
|
---|
| 194 | if (maxProb < y[i]) {
|
---|
| 195 | maxProb = y[i];
|
---|
| 196 | maxProbClassIndex = i;
|
---|
| 197 | }
|
---|
| 198 | }
|
---|
| 199 | yield return classValues[maxProbClassIndex];
|
---|
| 200 | }
|
---|
| 201 | }
|
---|
| 202 |
|
---|
[6603] | 203 | public INearestNeighbourRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
[8528] | 204 | return new NearestNeighbourRegressionSolution(new RegressionProblemData(problemData), this);
|
---|
[6603] | 205 | }
|
---|
| 206 | IRegressionSolution IRegressionModel.CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
| 207 | return CreateRegressionSolution(problemData);
|
---|
| 208 | }
|
---|
[6604] | 209 | public INearestNeighbourClassificationSolution CreateClassificationSolution(IClassificationProblemData problemData) {
|
---|
[8528] | 210 | return new NearestNeighbourClassificationSolution(new ClassificationProblemData(problemData), this);
|
---|
[6604] | 211 | }
|
---|
| 212 | IClassificationSolution IClassificationModel.CreateClassificationSolution(IClassificationProblemData problemData) {
|
---|
| 213 | return CreateClassificationSolution(problemData);
|
---|
| 214 | }
|
---|
[6603] | 215 |
|
---|
[6583] | 216 | #region events
|
---|
| 217 | public event EventHandler Changed;
|
---|
| 218 | private void OnChanged(EventArgs e) {
|
---|
| 219 | var handlers = Changed;
|
---|
| 220 | if (handlers != null)
|
---|
| 221 | handlers(this, e);
|
---|
| 222 | }
|
---|
| 223 | #endregion
|
---|
| 224 |
|
---|
| 225 | #region persistence
|
---|
[6584] | 226 | [Storable]
|
---|
| 227 | public double KDTreeApproxF {
|
---|
| 228 | get { return kdTree.approxf; }
|
---|
| 229 | set { kdTree.approxf = value; }
|
---|
| 230 | }
|
---|
| 231 | [Storable]
|
---|
| 232 | public double[] KDTreeBoxMax {
|
---|
| 233 | get { return kdTree.boxmax; }
|
---|
| 234 | set { kdTree.boxmax = value; }
|
---|
| 235 | }
|
---|
| 236 | [Storable]
|
---|
| 237 | public double[] KDTreeBoxMin {
|
---|
| 238 | get { return kdTree.boxmin; }
|
---|
| 239 | set { kdTree.boxmin = value; }
|
---|
| 240 | }
|
---|
| 241 | [Storable]
|
---|
| 242 | public double[] KDTreeBuf {
|
---|
| 243 | get { return kdTree.buf; }
|
---|
| 244 | set { kdTree.buf = value; }
|
---|
| 245 | }
|
---|
| 246 | [Storable]
|
---|
| 247 | public double[] KDTreeCurBoxMax {
|
---|
| 248 | get { return kdTree.curboxmax; }
|
---|
| 249 | set { kdTree.curboxmax = value; }
|
---|
| 250 | }
|
---|
| 251 | [Storable]
|
---|
| 252 | public double[] KDTreeCurBoxMin {
|
---|
| 253 | get { return kdTree.curboxmin; }
|
---|
| 254 | set { kdTree.curboxmin = value; }
|
---|
| 255 | }
|
---|
| 256 | [Storable]
|
---|
| 257 | public double KDTreeCurDist {
|
---|
| 258 | get { return kdTree.curdist; }
|
---|
| 259 | set { kdTree.curdist = value; }
|
---|
| 260 | }
|
---|
| 261 | [Storable]
|
---|
| 262 | public int KDTreeDebugCounter {
|
---|
| 263 | get { return kdTree.debugcounter; }
|
---|
| 264 | set { kdTree.debugcounter = value; }
|
---|
| 265 | }
|
---|
| 266 | [Storable]
|
---|
| 267 | public int[] KDTreeIdx {
|
---|
| 268 | get { return kdTree.idx; }
|
---|
| 269 | set { kdTree.idx = value; }
|
---|
| 270 | }
|
---|
| 271 | [Storable]
|
---|
| 272 | public int KDTreeKCur {
|
---|
| 273 | get { return kdTree.kcur; }
|
---|
| 274 | set { kdTree.kcur = value; }
|
---|
| 275 | }
|
---|
| 276 | [Storable]
|
---|
| 277 | public int KDTreeKNeeded {
|
---|
| 278 | get { return kdTree.kneeded; }
|
---|
| 279 | set { kdTree.kneeded = value; }
|
---|
| 280 | }
|
---|
| 281 | [Storable]
|
---|
| 282 | public int KDTreeN {
|
---|
| 283 | get { return kdTree.n; }
|
---|
| 284 | set { kdTree.n = value; }
|
---|
| 285 | }
|
---|
| 286 | [Storable]
|
---|
| 287 | public int[] KDTreeNodes {
|
---|
| 288 | get { return kdTree.nodes; }
|
---|
| 289 | set { kdTree.nodes = value; }
|
---|
| 290 | }
|
---|
| 291 | [Storable]
|
---|
| 292 | public int KDTreeNormType {
|
---|
| 293 | get { return kdTree.normtype; }
|
---|
| 294 | set { kdTree.normtype = value; }
|
---|
| 295 | }
|
---|
| 296 | [Storable]
|
---|
| 297 | public int KDTreeNX {
|
---|
| 298 | get { return kdTree.nx; }
|
---|
| 299 | set { kdTree.nx = value; }
|
---|
| 300 | }
|
---|
| 301 | [Storable]
|
---|
| 302 | public int KDTreeNY {
|
---|
| 303 | get { return kdTree.ny; }
|
---|
| 304 | set { kdTree.ny = value; }
|
---|
| 305 | }
|
---|
| 306 | [Storable]
|
---|
| 307 | public double[] KDTreeR {
|
---|
| 308 | get { return kdTree.r; }
|
---|
| 309 | set { kdTree.r = value; }
|
---|
| 310 | }
|
---|
| 311 | [Storable]
|
---|
| 312 | public double KDTreeRNeeded {
|
---|
| 313 | get { return kdTree.rneeded; }
|
---|
| 314 | set { kdTree.rneeded = value; }
|
---|
| 315 | }
|
---|
| 316 | [Storable]
|
---|
| 317 | public bool KDTreeSelfMatch {
|
---|
| 318 | get { return kdTree.selfmatch; }
|
---|
| 319 | set { kdTree.selfmatch = value; }
|
---|
| 320 | }
|
---|
| 321 | [Storable]
|
---|
| 322 | public double[] KDTreeSplits {
|
---|
| 323 | get { return kdTree.splits; }
|
---|
| 324 | set { kdTree.splits = value; }
|
---|
| 325 | }
|
---|
| 326 | [Storable]
|
---|
| 327 | public int[] KDTreeTags {
|
---|
| 328 | get { return kdTree.tags; }
|
---|
| 329 | set { kdTree.tags = value; }
|
---|
| 330 | }
|
---|
| 331 | [Storable]
|
---|
| 332 | public double[] KDTreeX {
|
---|
| 333 | get { return kdTree.x; }
|
---|
| 334 | set { kdTree.x = value; }
|
---|
| 335 | }
|
---|
| 336 | [Storable]
|
---|
| 337 | public double[,] KDTreeXY {
|
---|
| 338 | get { return kdTree.xy; }
|
---|
| 339 | set { kdTree.xy = value; }
|
---|
| 340 | }
|
---|
[6583] | 341 | #endregion
|
---|
| 342 | }
|
---|
| 343 | }
|
---|