[8606] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[12009] | 3 | * Copyright (C) 2002-2015 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[8606] | 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.Encodings.SymbolicExpressionTreeEncoding;
|
---|
| 28 | using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
|
---|
| 29 |
|
---|
| 30 | namespace HeuristicLab.Problems.DataAnalysis.Symbolic.Classification {
|
---|
| 31 | /// <summary>
|
---|
| 32 | /// Represents a nearest neighbour model for regression and classification
|
---|
| 33 | /// </summary>
|
---|
| 34 | [StorableClass]
|
---|
| 35 | [Item("SymbolicNearestNeighbourClassificationModel", "Represents a nearest neighbour model for symbolic classification.")]
|
---|
| 36 | public sealed class SymbolicNearestNeighbourClassificationModel : SymbolicClassificationModel {
|
---|
| 37 |
|
---|
| 38 | [Storable]
|
---|
| 39 | private int k;
|
---|
| 40 | [Storable]
|
---|
[9003] | 41 | private List<double> trainedClasses;
|
---|
[8978] | 42 | [Storable]
|
---|
[9003] | 43 | private List<double> trainedEstimatedValues;
|
---|
| 44 |
|
---|
| 45 | [Storable]
|
---|
[8978] | 46 | private ClassFrequencyComparer frequencyComparer;
|
---|
[8606] | 47 |
|
---|
| 48 | [StorableConstructor]
|
---|
| 49 | private SymbolicNearestNeighbourClassificationModel(bool deserializing) : base(deserializing) { }
|
---|
| 50 | private SymbolicNearestNeighbourClassificationModel(SymbolicNearestNeighbourClassificationModel original, Cloner cloner)
|
---|
| 51 | : base(original, cloner) {
|
---|
| 52 | k = original.k;
|
---|
[8978] | 53 | frequencyComparer = new ClassFrequencyComparer(original.frequencyComparer);
|
---|
[9003] | 54 | trainedEstimatedValues = new List<double>(original.trainedEstimatedValues);
|
---|
| 55 | trainedClasses = new List<double>(original.trainedClasses);
|
---|
[8606] | 56 | }
|
---|
[14027] | 57 | public SymbolicNearestNeighbourClassificationModel(string targetVariable, int k, ISymbolicExpressionTree tree, ISymbolicDataAnalysisExpressionTreeInterpreter interpreter, double lowerEstimationLimit = double.MinValue, double upperEstimationLimit = double.MaxValue)
|
---|
| 58 | : base(targetVariable, tree, interpreter, lowerEstimationLimit, upperEstimationLimit) {
|
---|
[8606] | 59 | this.k = k;
|
---|
[8978] | 60 | frequencyComparer = new ClassFrequencyComparer();
|
---|
[9003] | 61 |
|
---|
[8606] | 62 | }
|
---|
| 63 |
|
---|
| 64 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
| 65 | return new SymbolicNearestNeighbourClassificationModel(this, cloner);
|
---|
| 66 | }
|
---|
| 67 |
|
---|
[12702] | 68 | public override IEnumerable<double> GetEstimatedClassValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
[8978] | 69 | var estimatedValues = Interpreter.GetSymbolicExpressionTreeValues(SymbolicExpressionTree, dataset, rows)
|
---|
| 70 | .LimitToRange(LowerEstimationLimit, UpperEstimationLimit);
|
---|
[8606] | 71 | foreach (var ev in estimatedValues) {
|
---|
[9003] | 72 | // find the range [lower, upper[ of trainedTargetValues that contains the k closest neighbours
|
---|
| 73 | // the range can span more than k elements when there are equal estimated values
|
---|
| 74 |
|
---|
[8978] | 75 | // find the index of the training-point to which distance is shortest
|
---|
[9003] | 76 | int lower = trainedEstimatedValues.BinarySearch(ev);
|
---|
| 77 | int upper;
|
---|
| 78 | // if the element was not found exactly, BinarySearch returns the complement of the index of the next larger item
|
---|
| 79 | if (lower < 0) {
|
---|
| 80 | lower = ~lower;
|
---|
| 81 | // lower is not necessarily the closer one
|
---|
| 82 | // determine which element is closer to ev (lower - 1) or (lower)
|
---|
| 83 | if (lower == trainedEstimatedValues.Count ||
|
---|
| 84 | (lower > 0 && Math.Abs(ev - trainedEstimatedValues[lower - 1]) < Math.Abs(ev - trainedEstimatedValues[lower]))) {
|
---|
| 85 | lower = lower - 1;
|
---|
| 86 | }
|
---|
| 87 | }
|
---|
| 88 | upper = lower + 1;
|
---|
| 89 | // at this point we have a range [lower, upper[ that includes only the closest element to ev
|
---|
| 90 |
|
---|
| 91 | // expand the range to left or right looking for the nearest neighbors
|
---|
| 92 | while (upper - lower < Math.Min(k, trainedEstimatedValues.Count)) {
|
---|
| 93 | bool lowerIsCloser = upper >= trainedEstimatedValues.Count ||
|
---|
| 94 | (lower > 0 && ev - trainedEstimatedValues[lower] <= trainedEstimatedValues[upper] - ev);
|
---|
| 95 | bool upperIsCloser = lower <= 0 ||
|
---|
| 96 | (upper < trainedEstimatedValues.Count &&
|
---|
| 97 | ev - trainedEstimatedValues[lower] >= trainedEstimatedValues[upper] - ev);
|
---|
| 98 | if (!lowerIsCloser && !upperIsCloser) break;
|
---|
[9002] | 99 | if (lowerIsCloser) {
|
---|
[8606] | 100 | lower--;
|
---|
[9003] | 101 | // eat up all equal values
|
---|
| 102 | while (lower > 0 && trainedEstimatedValues[lower - 1].IsAlmost(trainedEstimatedValues[lower]))
|
---|
| 103 | lower--;
|
---|
[9002] | 104 | }
|
---|
| 105 | if (upperIsCloser) {
|
---|
[8606] | 106 | upper++;
|
---|
[9003] | 107 | while (upper < trainedEstimatedValues.Count &&
|
---|
| 108 | trainedEstimatedValues[upper - 1].IsAlmost(trainedEstimatedValues[upper]))
|
---|
| 109 | upper++;
|
---|
[8606] | 110 | }
|
---|
| 111 | }
|
---|
[8978] | 112 | // majority voting with preference for bigger class in case of tie
|
---|
[9003] | 113 | yield return Enumerable.Range(lower, upper - lower)
|
---|
| 114 | .Select(i => trainedClasses[i])
|
---|
| 115 | .GroupBy(c => c)
|
---|
| 116 | .Select(g => new { Class = g.Key, Votes = g.Count() })
|
---|
| 117 | .MaxItems(p => p.Votes)
|
---|
| 118 | .OrderByDescending(m => m.Class, frequencyComparer)
|
---|
| 119 | .First().Class;
|
---|
[8606] | 120 | }
|
---|
| 121 | }
|
---|
| 122 |
|
---|
| 123 | public override void RecalculateModelParameters(IClassificationProblemData problemData, IEnumerable<int> rows) {
|
---|
[8978] | 124 | var estimatedValues = Interpreter.GetSymbolicExpressionTreeValues(SymbolicExpressionTree, problemData.Dataset, rows)
|
---|
| 125 | .LimitToRange(LowerEstimationLimit, UpperEstimationLimit);
|
---|
[8606] | 126 | var targetValues = problemData.Dataset.GetDoubleValues(problemData.TargetVariable, rows);
|
---|
[9003] | 127 | var trainedClasses = targetValues.ToArray();
|
---|
| 128 | var trainedEstimatedValues = estimatedValues.ToArray();
|
---|
[8606] | 129 |
|
---|
[9003] | 130 | Array.Sort(trainedEstimatedValues, trainedClasses);
|
---|
| 131 | this.trainedClasses = new List<double>(trainedClasses);
|
---|
| 132 | this.trainedEstimatedValues = new List<double>(trainedEstimatedValues);
|
---|
[8606] | 133 |
|
---|
[9003] | 134 | var freq = trainedClasses
|
---|
| 135 | .GroupBy(c => c)
|
---|
| 136 | .ToDictionary(g => g.Key, g => g.Count());
|
---|
| 137 | this.frequencyComparer = new ClassFrequencyComparer(freq);
|
---|
[8606] | 138 | }
|
---|
| 139 |
|
---|
| 140 | public override ISymbolicClassificationSolution CreateClassificationSolution(IClassificationProblemData problemData) {
|
---|
[9002] | 141 | return new SymbolicClassificationSolution((ISymbolicClassificationModel)Clone(), problemData);
|
---|
[8606] | 142 | }
|
---|
| 143 | }
|
---|
[8978] | 144 |
|
---|
| 145 | [StorableClass]
|
---|
[8979] | 146 | internal sealed class ClassFrequencyComparer : IComparer<double> {
|
---|
[8978] | 147 | [Storable]
|
---|
[9002] | 148 | private readonly Dictionary<double, int> classFrequencies;
|
---|
[8978] | 149 |
|
---|
| 150 | [StorableConstructor]
|
---|
| 151 | private ClassFrequencyComparer(bool deserializing) { }
|
---|
| 152 | public ClassFrequencyComparer() {
|
---|
| 153 | classFrequencies = new Dictionary<double, int>();
|
---|
| 154 | }
|
---|
| 155 | public ClassFrequencyComparer(Dictionary<double, int> frequencies) {
|
---|
| 156 | classFrequencies = frequencies;
|
---|
| 157 | }
|
---|
| 158 | public ClassFrequencyComparer(ClassFrequencyComparer original) {
|
---|
| 159 | classFrequencies = new Dictionary<double, int>(original.classFrequencies);
|
---|
| 160 | }
|
---|
| 161 |
|
---|
| 162 | public int Compare(double x, double y) {
|
---|
| 163 | bool cx = classFrequencies.ContainsKey(x), cy = classFrequencies.ContainsKey(y);
|
---|
| 164 | if (cx && cy)
|
---|
| 165 | return classFrequencies[x].CompareTo(classFrequencies[y]);
|
---|
| 166 | if (cx) return 1;
|
---|
| 167 | return -1;
|
---|
| 168 | }
|
---|
| 169 | }
|
---|
[8606] | 170 | }
|
---|