1 | #region License Information
|
---|
2 | /* HeuristicLab
|
---|
3 | * Copyright (C) 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.IO;
|
---|
25 | using System.Linq;
|
---|
26 | using System.Text;
|
---|
27 | using HeuristicLab.Common;
|
---|
28 | using HeuristicLab.Core;
|
---|
29 | using HEAL.Attic;
|
---|
30 | using HeuristicLab.Problems.DataAnalysis;
|
---|
31 | using LibSVM;
|
---|
32 |
|
---|
33 | namespace HeuristicLab.Algorithms.DataAnalysis {
|
---|
34 | /// <summary>
|
---|
35 | /// Represents a support vector machine model.
|
---|
36 | /// </summary>
|
---|
37 | [StorableType("669B0522-A27B-4DA6-B655-27C374D88C95")]
|
---|
38 | [Item("SupportVectorMachineModel", "Represents a support vector machine model.")]
|
---|
39 | public sealed class SupportVectorMachineModel : ClassificationModel, ISupportVectorMachineModel {
|
---|
40 | public override IEnumerable<string> VariablesUsedForPrediction {
|
---|
41 | get { return allowedInputVariables; }
|
---|
42 | }
|
---|
43 |
|
---|
44 |
|
---|
45 | private svm_model model;
|
---|
46 | /// <summary>
|
---|
47 | /// Gets or sets the SVM model.
|
---|
48 | /// </summary>
|
---|
49 | public svm_model Model {
|
---|
50 | get { return model; }
|
---|
51 | set {
|
---|
52 | if (value != model) {
|
---|
53 | if (value == null) throw new ArgumentNullException();
|
---|
54 | model = value;
|
---|
55 | OnChanged(EventArgs.Empty);
|
---|
56 | }
|
---|
57 | }
|
---|
58 | }
|
---|
59 |
|
---|
60 | /// <summary>
|
---|
61 | /// Gets or sets the range transformation for the model.
|
---|
62 | /// </summary>
|
---|
63 | private RangeTransform rangeTransform;
|
---|
64 | public RangeTransform RangeTransform {
|
---|
65 | get { return rangeTransform; }
|
---|
66 | set {
|
---|
67 | if (value != rangeTransform) {
|
---|
68 | if (value == null) throw new ArgumentNullException();
|
---|
69 | rangeTransform = value;
|
---|
70 | OnChanged(EventArgs.Empty);
|
---|
71 | }
|
---|
72 | }
|
---|
73 | }
|
---|
74 |
|
---|
75 | public Dataset SupportVectors {
|
---|
76 | get {
|
---|
77 | var data = new double[Model.sv_coef.Length, allowedInputVariables.Count()];
|
---|
78 | for (int i = 0; i < Model.sv_coef.Length; i++) {
|
---|
79 | var sv = Model.SV[i];
|
---|
80 | for (int j = 0; j < sv.Length; j++) {
|
---|
81 | data[i, j] = sv[j].value;
|
---|
82 | }
|
---|
83 | }
|
---|
84 | return new Dataset(allowedInputVariables, data);
|
---|
85 | }
|
---|
86 | }
|
---|
87 |
|
---|
88 | [Storable]
|
---|
89 | private string[] allowedInputVariables;
|
---|
90 | [Storable]
|
---|
91 | private double[] classValues; // only for SVM classification models
|
---|
92 |
|
---|
93 | [StorableConstructor]
|
---|
94 | private SupportVectorMachineModel(StorableConstructorFlag _) : base(_) { }
|
---|
95 | private SupportVectorMachineModel(SupportVectorMachineModel original, Cloner cloner)
|
---|
96 | : base(original, cloner) {
|
---|
97 | // only using a shallow copy here! (gkronber)
|
---|
98 | this.model = original.model;
|
---|
99 | this.rangeTransform = original.rangeTransform;
|
---|
100 | this.allowedInputVariables = (string[])original.allowedInputVariables.Clone();
|
---|
101 | if (original.classValues != null)
|
---|
102 | this.classValues = (double[])original.classValues.Clone();
|
---|
103 | }
|
---|
104 | public SupportVectorMachineModel(svm_model model, RangeTransform rangeTransform, string targetVariable, IEnumerable<string> allowedInputVariables, IEnumerable<double> classValues)
|
---|
105 | : this(model, rangeTransform, targetVariable, allowedInputVariables) {
|
---|
106 | this.classValues = classValues.ToArray();
|
---|
107 | }
|
---|
108 | public SupportVectorMachineModel(svm_model model, RangeTransform rangeTransform, string targetVariable, IEnumerable<string> allowedInputVariables)
|
---|
109 | : base(targetVariable) {
|
---|
110 | this.name = ItemName;
|
---|
111 | this.description = ItemDescription;
|
---|
112 | this.model = model;
|
---|
113 | this.rangeTransform = rangeTransform;
|
---|
114 | this.allowedInputVariables = allowedInputVariables.ToArray();
|
---|
115 | }
|
---|
116 |
|
---|
117 | public override IDeepCloneable Clone(Cloner cloner) {
|
---|
118 | return new SupportVectorMachineModel(this, cloner);
|
---|
119 | }
|
---|
120 |
|
---|
121 | #region IRegressionModel Members
|
---|
122 | public IEnumerable<double> GetEstimatedValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
123 | return GetEstimatedValuesHelper(dataset, rows);
|
---|
124 | }
|
---|
125 | public IRegressionSolution CreateRegressionSolution(IRegressionProblemData problemData) {
|
---|
126 | return new SupportVectorRegressionSolution(this, new RegressionProblemData(problemData));
|
---|
127 | }
|
---|
128 |
|
---|
129 | public bool IsProblemDataCompatible(IRegressionProblemData problemData, out string errorMessage) {
|
---|
130 | return RegressionModel.IsProblemDataCompatible(this, problemData, out errorMessage);
|
---|
131 | }
|
---|
132 | #endregion
|
---|
133 |
|
---|
134 | public override bool IsProblemDataCompatible(IDataAnalysisProblemData problemData, out string errorMessage) {
|
---|
135 | if (problemData == null) throw new ArgumentNullException("problemData", "The provided problemData is null.");
|
---|
136 |
|
---|
137 | var regressionProblemData = problemData as IRegressionProblemData;
|
---|
138 | if (regressionProblemData != null)
|
---|
139 | return IsProblemDataCompatible(regressionProblemData, out errorMessage);
|
---|
140 |
|
---|
141 | var classificationProblemData = problemData as IClassificationProblemData;
|
---|
142 | if (classificationProblemData != null)
|
---|
143 | return IsProblemDataCompatible(classificationProblemData, out errorMessage);
|
---|
144 |
|
---|
145 | throw new ArgumentException("The problem data is not compatible with this SVM. Instead a " + problemData.GetType().GetPrettyName() + " was provided.", "problemData");
|
---|
146 | }
|
---|
147 |
|
---|
148 | #region IClassificationModel Members
|
---|
149 | public override IEnumerable<double> GetEstimatedClassValues(IDataset dataset, IEnumerable<int> rows) {
|
---|
150 | if (classValues == null) throw new NotSupportedException();
|
---|
151 | // return the original class value instead of the predicted value of the model
|
---|
152 | // svm classification only works for integer classes
|
---|
153 | foreach (var estimated in GetEstimatedValuesHelper(dataset, rows)) {
|
---|
154 | // find closest class
|
---|
155 | double bestDist = double.MaxValue;
|
---|
156 | double bestClass = -1;
|
---|
157 | for (int i = 0; i < classValues.Length; i++) {
|
---|
158 | double d = Math.Abs(estimated - classValues[i]);
|
---|
159 | if (d < bestDist) {
|
---|
160 | bestDist = d;
|
---|
161 | bestClass = classValues[i];
|
---|
162 | if (d.IsAlmost(0.0)) break; // exact match no need to look further
|
---|
163 | }
|
---|
164 | }
|
---|
165 | yield return bestClass;
|
---|
166 | }
|
---|
167 | }
|
---|
168 |
|
---|
169 | public override IClassificationSolution CreateClassificationSolution(IClassificationProblemData problemData) {
|
---|
170 | return new SupportVectorClassificationSolution(this, new ClassificationProblemData(problemData));
|
---|
171 | }
|
---|
172 | #endregion
|
---|
173 |
|
---|
174 | private IEnumerable<double> GetEstimatedValuesHelper(IDataset dataset, IEnumerable<int> rows) {
|
---|
175 | // calculate predictions for the currently requested rows
|
---|
176 | svm_problem problem = SupportVectorMachineUtil.CreateSvmProblem(dataset, allowedInputVariables, rows);
|
---|
177 | svm_problem scaledProblem = rangeTransform.Scale(problem);
|
---|
178 |
|
---|
179 | for (int i = 0; i < problem.l; i++) {
|
---|
180 | yield return svm.svm_predict(Model, scaledProblem.x[i]);
|
---|
181 | }
|
---|
182 | }
|
---|
183 |
|
---|
184 | #region events
|
---|
185 | public event EventHandler Changed;
|
---|
186 | private void OnChanged(EventArgs e) {
|
---|
187 | var handlers = Changed;
|
---|
188 | if (handlers != null)
|
---|
189 | handlers(this, e);
|
---|
190 | }
|
---|
191 | #endregion
|
---|
192 |
|
---|
193 | #region persistence
|
---|
194 | [Storable]
|
---|
195 | private string ModelAsString {
|
---|
196 | get {
|
---|
197 | using (MemoryStream stream = new MemoryStream()) {
|
---|
198 | svm.svm_save_model(new StreamWriter(stream), Model);
|
---|
199 | stream.Seek(0, System.IO.SeekOrigin.Begin);
|
---|
200 | StreamReader reader = new StreamReader(stream);
|
---|
201 | return reader.ReadToEnd();
|
---|
202 | }
|
---|
203 | }
|
---|
204 | set {
|
---|
205 | using (MemoryStream stream = new MemoryStream(Encoding.ASCII.GetBytes(value))) {
|
---|
206 | model = svm.svm_load_model(new StreamReader(stream));
|
---|
207 | }
|
---|
208 | }
|
---|
209 | }
|
---|
210 | [Storable]
|
---|
211 | private string RangeTransformAsString {
|
---|
212 | get {
|
---|
213 | using (MemoryStream stream = new MemoryStream()) {
|
---|
214 | RangeTransform.Write(stream, RangeTransform);
|
---|
215 | stream.Seek(0, System.IO.SeekOrigin.Begin);
|
---|
216 | StreamReader reader = new StreamReader(stream);
|
---|
217 | return reader.ReadToEnd();
|
---|
218 | }
|
---|
219 | }
|
---|
220 | set {
|
---|
221 | using (MemoryStream stream = new MemoryStream(Encoding.ASCII.GetBytes(value))) {
|
---|
222 | RangeTransform = RangeTransform.Read(stream);
|
---|
223 | }
|
---|
224 | }
|
---|
225 | }
|
---|
226 | #endregion
|
---|
227 | }
|
---|
228 | }
|
---|