source: trunk/sources/HeuristicLab.Algorithms.DataAnalysis/3.4/Linear/LinearRegression.cs @ 14826

Last change on this file since 14826 was 14826, checked in by gkronber, 6 months ago

#2650: merged the factors branch into trunk

File size: 6.7 KB
Line 
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
22using System;
23using System.Collections.Generic;
24using System.Linq;
25using System.Threading;
26using HeuristicLab.Common;
27using HeuristicLab.Core;
28using HeuristicLab.Data;
29using HeuristicLab.Encodings.SymbolicExpressionTreeEncoding;
30using HeuristicLab.Optimization;
31using HeuristicLab.Persistence.Default.CompositeSerializers.Storable;
32using HeuristicLab.Problems.DataAnalysis;
33using HeuristicLab.Problems.DataAnalysis.Symbolic;
34using HeuristicLab.Problems.DataAnalysis.Symbolic.Regression;
35
36namespace HeuristicLab.Algorithms.DataAnalysis {
37  /// <summary>
38  /// Linear regression data analysis algorithm.
39  /// </summary>
40  [Item("Linear Regression (LR)", "Linear regression data analysis algorithm (wrapper for ALGLIB).")]
41  [Creatable(CreatableAttribute.Categories.DataAnalysisRegression, Priority = 100)]
42  [StorableClass]
43  public sealed class LinearRegression : FixedDataAnalysisAlgorithm<IRegressionProblem> {
44    private const string LinearRegressionModelResultName = "Linear regression solution";
45
46    [StorableConstructor]
47    private LinearRegression(bool deserializing) : base(deserializing) { }
48    private LinearRegression(LinearRegression original, Cloner cloner)
49      : base(original, cloner) {
50    }
51    public LinearRegression()
52      : base() {
53      Problem = new RegressionProblem();
54    }
55    [StorableHook(HookType.AfterDeserialization)]
56    private void AfterDeserialization() { }
57
58    public override IDeepCloneable Clone(Cloner cloner) {
59      return new LinearRegression(this, cloner);
60    }
61
62    #region linear regression
63    protected override void Run(CancellationToken cancellationToken) {
64      double rmsError, cvRmsError;
65      var solution = CreateLinearRegressionSolution(Problem.ProblemData, out rmsError, out cvRmsError);
66      Results.Add(new Result(LinearRegressionModelResultName, "The linear regression solution.", solution));
67      Results.Add(new Result("Root mean square error", "The root of the mean of squared errors of the linear regression solution on the training set.", new DoubleValue(rmsError)));
68      Results.Add(new Result("Estimated root mean square error (cross-validation)", "The estimated root of the mean of squared errors of the linear regression solution via cross validation.", new DoubleValue(cvRmsError)));
69    }
70
71    public static ISymbolicRegressionSolution CreateLinearRegressionSolution(IRegressionProblemData problemData, out double rmsError, out double cvRmsError) {
72      var dataset = problemData.Dataset;
73      string targetVariable = problemData.TargetVariable;
74      IEnumerable<string> allowedInputVariables = problemData.AllowedInputVariables;
75      IEnumerable<int> rows = problemData.TrainingIndices;
76      var doubleVariables = allowedInputVariables.Where(dataset.VariableHasType<double>);
77      var factorVariableNames = allowedInputVariables.Where(dataset.VariableHasType<string>);
78      var factorVariables = AlglibUtil.GetFactorVariableValues(dataset, factorVariableNames, rows);
79      double[,] binaryMatrix = AlglibUtil.PrepareInputMatrix(dataset, factorVariables, rows);
80      double[,] doubleVarMatrix = AlglibUtil.PrepareInputMatrix(dataset, doubleVariables.Concat(new string[] { targetVariable }), rows);
81      var inputMatrix = binaryMatrix.HorzCat(doubleVarMatrix);
82
83      if (inputMatrix.Cast<double>().Any(x => double.IsNaN(x) || double.IsInfinity(x)))
84        throw new NotSupportedException("Linear regression does not support NaN or infinity values in the input dataset.");
85
86      alglib.linearmodel lm = new alglib.linearmodel();
87      alglib.lrreport ar = new alglib.lrreport();
88      int nRows = inputMatrix.GetLength(0);
89      int nFeatures = inputMatrix.GetLength(1) - 1;
90      double[] coefficients = new double[nFeatures + 1]; // last coefficient is for the constant
91
92      int retVal = 1;
93      alglib.lrbuild(inputMatrix, nRows, nFeatures, out retVal, out lm, out ar);
94      if (retVal != 1) throw new ArgumentException("Error in calculation of linear regression solution");
95      rmsError = ar.rmserror;
96      cvRmsError = ar.cvrmserror;
97
98      alglib.lrunpack(lm, out coefficients, out nFeatures);
99
100      ISymbolicExpressionTree tree = new SymbolicExpressionTree(new ProgramRootSymbol().CreateTreeNode());
101      ISymbolicExpressionTreeNode startNode = new StartSymbol().CreateTreeNode();
102      tree.Root.AddSubtree(startNode);
103      ISymbolicExpressionTreeNode addition = new Addition().CreateTreeNode();
104      startNode.AddSubtree(addition);
105
106      int col = 0;
107      foreach (var kvp in factorVariables) {
108        var varName = kvp.Key;
109        foreach (var cat in kvp.Value) {
110          BinaryFactorVariableTreeNode vNode =
111            (BinaryFactorVariableTreeNode)new HeuristicLab.Problems.DataAnalysis.Symbolic.BinaryFactorVariable().CreateTreeNode();
112          vNode.VariableName = varName;
113          vNode.VariableValue = cat;
114          vNode.Weight = coefficients[col];
115          addition.AddSubtree(vNode);
116          col++;
117        }
118      }
119      foreach (string column in doubleVariables) {
120        VariableTreeNode vNode = (VariableTreeNode)new HeuristicLab.Problems.DataAnalysis.Symbolic.Variable().CreateTreeNode();
121        vNode.VariableName = column;
122        vNode.Weight = coefficients[col];
123        addition.AddSubtree(vNode);
124        col++;
125      }
126
127      ConstantTreeNode cNode = (ConstantTreeNode)new Constant().CreateTreeNode();
128      cNode.Value = coefficients[coefficients.Length - 1];
129      addition.AddSubtree(cNode);
130
131      SymbolicRegressionSolution solution = new SymbolicRegressionSolution(new SymbolicRegressionModel(problemData.TargetVariable, tree, new SymbolicDataAnalysisExpressionTreeLinearInterpreter()), (IRegressionProblemData)problemData.Clone());
132      solution.Model.Name = "Linear Regression Model";
133      solution.Name = "Linear Regression Solution";
134      return solution;
135    }
136    #endregion
137  }
138}
Note: See TracBrowser for help on using the repository browser.