1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 2002-2016 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;
|
---|
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 |
|
---|
29 | namespace HeuristicLab.Problems.DataAnalysis {
|
---|
30 | /// <summary>
|
---|
31 | /// Represents regression solutions that contain an ensemble of multiple regression models
|
---|
32 | /// </summary>
|
---|
33 | [StorableClass]
|
---|
34 | [Item("Regression Ensemble Model", "A regression model that contains an ensemble of multiple regression models")]
|
---|
35 | public sealed class RegressionEnsembleModel : RegressionModel, IRegressionEnsembleModel {
|
---|
36 | public override IEnumerable<string> VariablesUsedForPrediction {
|
---|
37 | get { return models.SelectMany(x => x.VariablesUsedForPrediction).Distinct().OrderBy(x => x); }
|
---|
38 | }
|
---|
39 |
|
---|
40 | private List<IRegressionModel> models;
|
---|
41 | public IEnumerable<IRegressionModel> Models {
|
---|
42 | get { return new List<IRegressionModel>(models); }
|
---|
43 | }
|
---|
44 |
|
---|
45 | [Storable(Name = "Models")]
|
---|
46 | private IEnumerable<IRegressionModel> StorableModels {
|
---|
47 | get { return models; }
|
---|
48 | set { models = value.ToList(); }
|
---|
49 | }
|
---|
50 |
|
---|
51 | private List<double> modelWeights;
|
---|
52 | public IEnumerable<double> ModelWeights {
|
---|
53 | get { return modelWeights; }
|
---|
54 | }
|
---|
55 |
|
---|
56 | [Storable(Name = "ModelWeights")]
|
---|
57 | private IEnumerable<double> StorableModelWeights {
|
---|
58 | get { return modelWeights; }
|
---|
59 | set { modelWeights = value.ToList(); }
|
---|
60 | }
|
---|
61 |
|
---|
62 | [Storable]
|
---|
63 | private bool averageModelEstimates = true;
|
---|
64 | public bool AverageModelEstimates {
|
---|
65 | get { return averageModelEstimates; }
|
---|
66 | set {
|
---|
67 | if (averageModelEstimates != value) {
|
---|
68 | averageModelEstimates = value;
|
---|
69 | OnChanged();
|
---|
70 | }
|
---|
71 | }
|
---|
72 | }
|
---|
73 |
|
---|
74 | #region backwards compatiblity 3.3.5
|
---|
75 | [Storable(Name = "models", AllowOneWay = true)]
|
---|
76 | private List<IRegressionModel> OldStorableModels {
|
---|
77 | set { models = value; }
|
---|
78 | }
|
---|
79 | #endregion
|
---|
80 |
|
---|
81 | [StorableHook(HookType.AfterDeserialization)]
|
---|
82 | private void AfterDeserialization() {
|
---|
83 | // BackwardsCompatibility 3.3.14
|
---|
84 | #region Backwards compatible code, remove with 3.4
|
---|
85 | if (modelWeights == null || !modelWeights.Any())
|
---|
86 | modelWeights = new List<double>(models.Select(m => 1.0));
|
---|
87 | #endregion
|
---|
88 | }
|
---|
89 |
|
---|
90 | [StorableConstructor]
|
---|
91 | private RegressionEnsembleModel(bool deserializing) : base(deserializing) { }
|
---|
92 | private RegressionEnsembleModel(RegressionEnsembleModel original, Cloner cloner)
|
---|
93 | : base(original, cloner) {
|
---|
94 | this.models = original.Models.Select(cloner.Clone).ToList();
|
---|
95 | this.modelWeights = new List<double>(original.ModelWeights);
|
---|
96 | this.averageModelEstimates = original.averageModelEstimates;
|
---|
97 | }
|
---|
98 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
99 | return new RegressionEnsembleModel(this, cloner);
|
---|
100 | }
|
---|
101 |
|
---|
102 | public RegressionEnsembleModel() : this(Enumerable.Empty<IRegressionModel>()) { }
|
---|
103 | public RegressionEnsembleModel(IEnumerable<IRegressionModel> models) : this(models, models.Select(m => 1.0)) { }
|
---|
104 | public RegressionEnsembleModel(IEnumerable<IRegressionModel> models, IEnumerable<double> modelWeights)
|
---|
105 | : base(string.Empty) {
|
---|
106 | this.name = ItemName;
|
---|
107 | this.description = ItemDescription;
|
---|
108 |
|
---|
109 | this.models = new List<IRegressionModel>(models);
|
---|
110 | this.modelWeights = new List<double>(modelWeights);
|
---|
111 |
|
---|
112 | if (this.models.Any()) this.TargetVariable = this.models.First().TargetVariable;
|
---|
113 | }
|
---|
114 |
|
---|
115 | public void Add(IRegressionModel model) {
|
---|
116 | if (string.IsNullOrEmpty(TargetVariable)) TargetVariable = model.TargetVariable;
|
---|
117 | Add(model, 1.0);
|
---|
118 | }
|
---|
119 | public void Add(IRegressionModel model, double weight) {
|
---|
120 | if (string.IsNullOrEmpty(TargetVariable)) TargetVariable = model.TargetVariable;
|
---|
121 |
|
---|
122 | models.Add(model);
|
---|
123 | modelWeights.Add(weight);
|
---|
124 | OnChanged();
|
---|
125 | }
|
---|
126 |
|
---|
127 | public void AddRange(IEnumerable<IRegressionModel> models) {
|
---|
128 | AddRange(models, models.Select(m => 1.0));
|
---|
129 | }
|
---|
130 | public void AddRange(IEnumerable<IRegressionModel> models, IEnumerable<double> weights) {
|
---|
131 | if (string.IsNullOrEmpty(TargetVariable)) TargetVariable = models.First().TargetVariable;
|
---|
132 |
|
---|
133 | this.models.AddRange(models);
|
---|
134 | modelWeights.AddRange(weights);
|
---|
135 | OnChanged();
|
---|
136 | }
|
---|
137 |
|
---|
138 | public void Remove(IRegressionModel model) {
|
---|
139 | var index = models.IndexOf(model);
|
---|
140 | models.RemoveAt(index);
|
---|
141 | modelWeights.RemoveAt(index);
|
---|
142 |
|
---|
143 | if (!models.Any()) TargetVariable = string.Empty;
|
---|
144 | OnChanged();
|
---|
145 | }
|
---|
146 | public void RemoveRange(IEnumerable<IRegressionModel> models) {
|
---|
147 | foreach (var model in models) {
|
---|
148 | var index = this.models.IndexOf(model);
|
---|
149 | this.models.RemoveAt(index);
|
---|
150 | modelWeights.RemoveAt(index);
|
---|
151 | }
|
---|
152 |
|
---|
153 | if (!models.Any()) TargetVariable = string.Empty;
|
---|
154 | OnChanged();
|
---|
155 | }
|
---|
156 |
|
---|
157 | public double GetModelWeight(IRegressionModel model) {
|
---|
158 | var index = models.IndexOf(model);
|
---|
159 | return modelWeights[index];
|
---|
160 | }
|
---|
161 | public void SetModelWeight(IRegressionModel model, double weight) {
|
---|
162 | var index = models.IndexOf(model);
|
---|
163 | modelWeights[index] = weight;
|
---|
164 | OnChanged();
|
---|
165 | }
|
---|
166 |
|
---|
167 | #region evaluation
|
---|
168 | public IEnumerable<IEnumerable<double>> GetEstimatedValueVectors(IDataset dataset, IEnumerable<int> rows) {
|
---|
169 | var estimatedValuesEnumerators = (from model in models
|
---|
170 | let weight = GetModelWeight(model)
|
---|
171 | select model.GetEstimatedValues(dataset, rows).Select(e => weight * e)
|
---|
172 | .GetEnumerator()).ToList();
|
---|
173 |
|
---|
174 | while (estimatedValuesEnumerators.All(en => en.MoveNext())) {
|
---|
175 | yield return from enumerator in estimatedValuesEnumerators
|
---|
176 | select enumerator.Current;
|
---|
177 | }
|
---|
178 | }
|
---|
179 |
|
---|
180 | public override IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
181 | double weightsSum = modelWeights.Sum();
|
---|
182 | var summedEstimates = from estimatedValuesVector in GetEstimatedValueVectors(dataset, rows)
|
---|
183 | select estimatedValuesVector.DefaultIfEmpty(double.NaN).Sum();
|
---|
184 |
|
---|
185 | if (AverageModelEstimates)
|
---|
186 | return summedEstimates.Select(v => v / weightsSum);
|
---|
187 | else
|
---|
188 | return summedEstimates;
|
---|
189 |
|
---|
190 | }
|
---|
191 |
|
---|
192 | public IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows, Func<int, IRegressionModel, bool> modelSelectionPredicate) {
|
---|
193 | var estimatedValuesEnumerators = GetEstimatedValueVectors(dataset, rows).GetEnumerator();
|
---|
194 | var rowsEnumerator = rows.GetEnumerator();
|
---|
195 |
|
---|
196 | while (rowsEnumerator.MoveNext() & estimatedValuesEnumerators.MoveNext()) {
|
---|
197 | var estimatedValueEnumerator = estimatedValuesEnumerators.Current.GetEnumerator();
|
---|
198 | int currentRow = rowsEnumerator.Current;
|
---|
199 | double weightsSum = 0.0;
|
---|
200 | double filteredEstimatesSum = 0.0;
|
---|
201 |
|
---|
202 | for (int m = 0; m < models.Count; m++) {
|
---|
203 | estimatedValueEnumerator.MoveNext();
|
---|
204 | var model = models[m];
|
---|
205 | if (!modelSelectionPredicate(currentRow, model)) continue;
|
---|
206 |
|
---|
207 | filteredEstimatesSum += estimatedValueEnumerator.Current;
|
---|
208 | weightsSum += modelWeights[m];
|
---|
209 | }
|
---|
210 |
|
---|
211 | if (AverageModelEstimates)
|
---|
212 | yield return filteredEstimatesSum / weightsSum;
|
---|
213 | else
|
---|
214 | yield return filteredEstimatesSum;
|
---|
215 | }
|
---|
216 | }
|
---|
217 |
|
---|
218 | #endregion
|
---|
219 |
|
---|
220 | public event EventHandler Changed;
|
---|
221 | private void OnChanged() {
|
---|
222 | var handler = Changed;
|
---|
223 | if (handler != null)
|
---|
224 | handler(this, EventArgs.Empty);
|
---|
225 | }
|
---|
226 |
|
---|
227 |
|
---|
228 | public override IRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
229 | return new RegressionEnsembleSolution(this, new RegressionEnsembleProblemData(problemData));
|
---|
230 | }
|
---|
231 | }
|
---|
232 | }
|
---|