source: branches/ClassificationEnsembleVoting/HeuristicLab.Problems.DataAnalysis.Views/3.4/Classification/ClassificationEnsembleSolutionAccuracyToCoveredSamples.cs @ 8423

#region License Information
/* HeuristicLab
 * Copyright (C) 2002-2012 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.Windows.Forms;
using System.Windows.Forms.DataVisualization.Charting;
using HeuristicLab.Data;
using HeuristicLab.MainForm;
using HeuristicLab.Problems.DataAnalysis.Interfaces;

namespace HeuristicLab.Problems.DataAnalysis.Views {
  [View("Accuracy Covered Dependence")]
  [Content(typeof(IClassificationEnsembleSolution))]
  public partial class ClassificationEnsembleSolutionAccuracyToCoveredSamples : DataAnalysisSolutionEvaluationView {
    private const string ACCURACYCOVERED = "Accuracy to Covered percentage";

    private const string SamplesComboBoxAllSamples = "All Samples";
    private const string SamplesComboBoxTrainingSamples = "Training Samples";
    private const string SamplesComboBoxTestSamples = "Test Samples";

    // zero is also a point
    private const int maxPoints = 101;

    public new ClassificationEnsembleSolution Content {
      get { return (ClassificationEnsembleSolution)base.Content; }
      set { base.Content = value; }
    }

    public ClassificationEnsembleSolutionAccuracyToCoveredSamples()
      : base() {
      InitializeComponent();

      SamplesComboBox.Items.AddRange(new string[] { SamplesComboBoxAllSamples, SamplesComboBoxTrainingSamples, SamplesComboBoxTestSamples });
      SamplesComboBox.SelectedIndex = 0;
      //configure axis
      this.chart.CustomizeAllChartAreas();
      this.chart.ChartAreas[0].CursorX.IsUserSelectionEnabled = true;
      this.chart.ChartAreas[0].AxisX.ScaleView.Zoomable = true;
      this.chart.ChartAreas[0].AxisX.IsStartedFromZero = true;
      this.chart.ChartAreas[0].AxisX.Minimum = 0;
      this.chart.ChartAreas[0].AxisX.Maximum = 1;
      this.chart.ChartAreas[0].AxisX.Title = "Covered Samples in %";

      this.chart.ChartAreas[0].CursorY.IsUserSelectionEnabled = true;
      this.chart.ChartAreas[0].AxisY.ScaleView.Zoomable = true;
      this.chart.ChartAreas[0].AxisY.IsStartedFromZero = true;
      this.chart.ChartAreas[0].AxisY.Minimum = 0;
      this.chart.ChartAreas[0].AxisY.Maximum = 1;
      this.chart.ChartAreas[0].AxisY.Title = "Accuracy";
    }

    private void RedrawChart() {
      this.chart.Series.Clear();
      if (Content != null) {

        double[] accuracy = new double[maxPoints + 1];
        double[] covered = new double[maxPoints + 1];

        IClassificationEnsembleSolutionWeightCalculator weightCalc = Content.WeightCalculator;
        var solutions = Content.ClassificationSolutions;
        double[] estimatedClassValues;
        double[] classValues;
        OnlineAccuracyCalculator accuracyCalc = new OnlineAccuracyCalculator();

        int rows;
        double[] confidences;

        if (SamplesComboBox.SelectedItem.ToString().Equals(SamplesComboBoxAllSamples)) {
          rows = Content.ProblemData.Dataset.Rows;
          estimatedClassValues = Content.EstimatedClassValues.ToArray();
          classValues = Content.ProblemData.Dataset.GetDoubleValues(Content.ProblemData.TargetVariable).ToArray();
          confidences = weightCalc.GetConfidence(solutions, Enumerable.Range(0, rows), estimatedClassValues).ToArray();
        } else {
          IntRange range;
          if (SamplesComboBox.SelectedItem.ToString().Equals(SamplesComboBoxTrainingSamples)) {
            range = Content.ProblemData.TrainingPartition;
            estimatedClassValues = Content.EstimatedTrainingClassValues.ToArray();
          } else if (SamplesComboBox.SelectedItem.ToString().Equals(SamplesComboBoxTestSamples)) {
            range = Content.ProblemData.TestPartition;
            estimatedClassValues = Content.EstimatedTestClassValues.ToArray();
          } else {
            return;
          }
          rows = range.End - range.Start;
          classValues = Content.ProblemData.Dataset.GetDoubleValues(Content.ProblemData.TargetVariable)
                                      .Skip(range.Start).Take(range.End - range.Start).ToArray();
          confidences = new double[rows];
          int index;
          for (int i = 0; i < rows; i++) {
            index = range.Start + i;
            confidences[i] = weightCalc.GetConfidence(GetRelevantSolutions(SamplesComboBox.SelectedItem.ToString(), solutions, index),
                                                      index, estimatedClassValues[i]);
          }
        }

        for (int i = 0; i < maxPoints; i++) {
          double confidenceValue = (1.0 / (maxPoints - 1)) * i;
          int notCovered = 0;

          for (int j = 0; j < rows; j++) {
            if (confidences[j] >= confidenceValue) {
              accuracyCalc.Add(classValues[j], estimatedClassValues[j]);
            } else {
              notCovered++;
            }
          }

          accuracy[i + 1] = accuracyCalc.Accuracy;
          covered[i] = 1.0 - (double)notCovered / (double)rows;
          accuracyCalc.Reset();
        }

        accuracy[0] = accuracy[1];
        covered[maxPoints] = 0.0;

        accuracy = accuracy.Reverse().ToArray();
        covered = covered.Reverse().ToArray();



        Series serie = this.chart.Series.Add(ACCURACYCOVERED);
        serie.LegendText = ACCURACYCOVERED;
        serie.ChartType = SeriesChartType.StepLine;
        //serie.MarkerStyle = MarkerStyle.Diamond;
        //serie.MarkerSize = 5;
        serie.Points.DataBindXY(covered, accuracy);
      }
    }

    protected IEnumerable<IClassificationSolution> GetRelevantSolutions(string samplesSelection, IEnumerable<IClassificationSolution> solutions, int curRow) {
      if (samplesSelection == SamplesComboBoxAllSamples)
        return solutions;
      else if (samplesSelection == SamplesComboBoxTrainingSamples)
        return solutions.Where(s => s.ProblemData.IsTrainingSample(curRow));
      else if (samplesSelection == SamplesComboBoxTestSamples)
        return solutions.Where(s => s.ProblemData.IsTestSample(curRow));
      else
        return new List<IClassificationSolution>();
    }

    #region events
    protected override void RegisterContentEvents() {
      base.RegisterContentEvents();
      Content.ModelChanged += new EventHandler(Content_ModelChanged);
      Content.ProblemDataChanged += new EventHandler(Content_ProblemDataChanged);
    }
    protected override void DeregisterContentEvents() {
      base.DeregisterContentEvents();
      Content.ModelChanged -= new EventHandler(Content_ModelChanged);
      Content.ProblemDataChanged -= new EventHandler(Content_ProblemDataChanged);
    }

    protected override void OnContentChanged() {
      base.OnContentChanged();
      RedrawChart();
    }
    private void Content_ProblemDataChanged(object sender, EventArgs e) {
      RedrawChart();
    }
    private void Content_ModelChanged(object sender, EventArgs e) {
      RedrawChart();
    }
    private void SamplesComboBox_SelectedIndexChanged(object sender, EventArgs e) {
      RedrawChart();
    }
    #endregion
  }
}