source: branches/2990_VariableImpactBasedFeatureSelection/HeuristicLab.Algorithms.DataAnalysis/3.4/FeatureSelection/VariableImpactBasedFeatureSelectionAlgorithm.cs @ 16705

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2019 Heuristic and Evolutionary Algorithms Laboratory (HEAL)
 *
 * This file is part of HeuristicLab.
 *
 * HeuristicLab is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 3 of the License, or
 * (at your option) any later version.
 *
 * HeuristicLab is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with HeuristicLab. If not, see <http://www.gnu.org/licenses/>.
 */
#endregion

using System;
using System.Collections.Generic;
using System.Linq;
using System.Threading;
using HeuristicLab.Algorithms.DataAnalysis;
using HeuristicLab.Analysis;
using HeuristicLab.Common;
using HeuristicLab.Core;
using HeuristicLab.Data;
using HeuristicLab.Optimization;
using HeuristicLab.Parameters;
using HEAL.Attic;

namespace HeuristicLab.Problems.DataAnalysis.FeatureSelection {
  [Item("VariableImpactBasedFeatureSelectionAlgorithm", "")]
  [Creatable(CreatableAttribute.Categories.DataAnalysisRegression, Priority = 9999)]
  [StorableType("EB47CA07-6F01-4FC1-9351-54ACB1F2CF24")]
  public class VariableImpactBasedFeatureSelectionAlgorithm : BasicAlgorithm {

    public override bool SupportsPause {
      get { return false; }
    }

    #region Problem Type
    public override Type ProblemType {
      get { return typeof(IRegressionProblem); }
    }
    public new IRegressionProblem Problem {
      get { return (IRegressionProblem)base.Problem; }
      set { base.Problem = value; }
    }
    #endregion

    #region Parameter Properties
    private ValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>> AlgorithmParameter {
      get { return (ValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>>)Parameters["Algorithm"]; }
    }

    private FixedValueParameter<PercentValue> FeaturesDropParameter {
      get { return (FixedValueParameter<PercentValue>)Parameters["FeaturesDrop"]; }
    }
    #endregion

    #region Results Parameter

    #endregion

    #region Constructor, Cloning & Persistence
    public VariableImpactBasedFeatureSelectionAlgorithm() {
      Parameters.Add(new ValueParameter<FixedDataAnalysisAlgorithm<IRegressionProblem>>("Algorithm", new RandomForestRegression()));
      Parameters.Add(new FixedValueParameter<PercentValue>("FeaturesDrop", new PercentValue(0.2)));

      // ToDo: Use ResultParameters
      //Parameters.Add(new ResultParameter<>());

      Problem = new RegressionProblem();
    }

    [StorableConstructor]
    protected VariableImpactBasedFeatureSelectionAlgorithm(StorableConstructorFlag _)
      : base(_) { }

    public VariableImpactBasedFeatureSelectionAlgorithm(VariableImpactBasedFeatureSelectionAlgorithm original, Cloner cloner)
      : base(original, cloner) { }
    public override IDeepCloneable Clone(Cloner cloner) {
      return new VariableImpactBasedFeatureSelectionAlgorithm(this, cloner);
    }
    #endregion

    protected override void Run(CancellationToken cancellationToken) {
      var clonedAlgorithm = (FixedDataAnalysisAlgorithm<IRegressionProblem>)AlgorithmParameter.Value.Clone();
      var clonedProblem = (IRegressionProblem)Problem.Clone();
      double featureDrop = FeaturesDropParameter.Value.Value;

      SetupAlgorithm(clonedAlgorithm, clonedProblem);
      Results.Add(new Result("Algorithm", clonedAlgorithm));

      var remainingFeatures = clonedProblem.ProblemData.InputVariables.CheckedItems.Select(x => x.Value.Value).ToList();

      var variableImpactsDataTable = new DataTable("VariableImpacts");
      foreach (var variable in clonedProblem.ProblemData.InputVariables) {
        variableImpactsDataTable.Rows.Add(new DataRow(variable.Value));
      }
      Results.Add(new Result("VariableImpacts", variableImpactsDataTable));

      var selectedVariablesResult = new StringMatrix(remainingFeatures.Count, 1) {
        ColumnNames = new[] { "StartUp" }
      };
      for (int i = 0; i < remainingFeatures.Count; i++) {
        selectedVariablesResult[i, 0] = remainingFeatures[i];
      }
      Results.Add(new Result("SelectedFeatures", selectedVariablesResult));

      var qualitiesResult = new DataTable("SolutionQualities");
      qualitiesResult.Rows.Add(new DataRow("MAE Training"));
      qualitiesResult.Rows.Add(new DataRow("MAE Test"));
      qualitiesResult.Rows.Add(new DataRow("R² Training") { VisualProperties = { SecondYAxis = true } });
      qualitiesResult.Rows.Add(new DataRow("R² Test") { VisualProperties = { SecondYAxis = true } });
      Results.Add(new Result("Qualities", qualitiesResult));


      int iteration = 0;
      while (remainingFeatures.Any()) {
        clonedAlgorithm.Start(cancellationToken);

        int numberOfRemainingVariables = (int)(remainingFeatures.Count * (1.0 - featureDrop)); // floor to avoid getting stuck
        var variableImpacts = GetVariableImpacts(clonedAlgorithm.Results).ToDictionary(x => x.Item1, x => x.Item2);

        remainingFeatures = variableImpacts
          .OrderByDescending(x => x.Value)
          .Take(numberOfRemainingVariables)
          .Select(x => x.Key)
          .ToList();

        foreach (var row in variableImpactsDataTable.Rows) {
          row.Values.Add(variableImpacts.ContainsKey(row.Name) ? variableImpacts[row.Name] : double.NaN);
        }

        ((IStringConvertibleMatrix)selectedVariablesResult).Columns++;
        selectedVariablesResult.ColumnNames = selectedVariablesResult.ColumnNames.Select(c => c.Replace("Column", "Iteration"));
        //selectedVariablesResult.ColumnNames = selectedVariablesResult.ColumnNames.Concat(new[] { $"Iteration {iteration}" });
        for (int i = 0; i < remainingFeatures.Count; i++) {
          selectedVariablesResult[i, selectedVariablesResult.Columns - 1] = remainingFeatures[i];
        }

        var solution = clonedAlgorithm.Results.Select(r => r.Value).OfType<IRegressionSolution>().Single();
        qualitiesResult.Rows["MAE Training"].Values.Add(solution.TrainingMeanAbsoluteError);
        qualitiesResult.Rows["MAE Test"].Values.Add(solution.TestMeanAbsoluteError);
        qualitiesResult.Rows["R² Training"].Values.Add(solution.TrainingRSquared);
        qualitiesResult.Rows["R² Test"].Values.Add(solution.TestRSquared);


        UpdateSelectedInputs(clonedProblem, remainingFeatures);

        iteration++;
      }
    }

    private static void SetupAlgorithm(FixedDataAnalysisAlgorithm<IRegressionProblem> algorithm, IRegressionProblem problem) {
      algorithm.Problem = problem;
      algorithm.Prepare(clearRuns: true);
    }

    private static IEnumerable<Tuple<string, double>> GetVariableImpacts(ResultCollection results) {
      //var solution = (IRegressionSolution)results["Random forest regression solution"].Value;
      var solution = results.Select(r => r.Value).OfType<IRegressionSolution>().Single();
      return RegressionSolutionVariableImpactsCalculator.CalculateImpacts(
        solution,
        replacementMethod: RegressionSolutionVariableImpactsCalculator.ReplacementMethodEnum.Shuffle,
        factorReplacementMethod: RegressionSolutionVariableImpactsCalculator.FactorReplacementMethodEnum.Shuffle,
        dataPartition: RegressionSolutionVariableImpactsCalculator.DataPartitionEnum.Training);
    }

    private void UpdateSelectedInputs(IRegressionProblem problem, List<string> remainingFeatures) {
      foreach (var inputFeature in problem.ProblemData.InputVariables) {
        bool isRemaining = remainingFeatures.Contains(inputFeature.Value);
        problem.ProblemData.InputVariables.SetItemCheckedState(inputFeature, isRemaining);
      }
    }
  }
}