source: branches/1776_ClassificationEnsembleVoting/HeuristicLab.Problems.DataAnalysis.Views/3.4/Classification/ClassificationEnsembleSolutionAccuracyToCoveredSamples.cs @ 16724

#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.Drawing;
using System.Linq;
using System.Windows.Forms;
using System.Windows.Forms.DataVisualization.Charting;
using HeuristicLab.MainForm;

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 AREA = "Area";

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

    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";

      AUCLabel.Parent = chart;
      AUCLabel.BackColor = Color.Transparent;
    }

    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 = null;
        double[] target;
        OnlineAccuracyCalculator accuracyCalc = new OnlineAccuracyCalculator();

        int[] indizes;
        double[] confidences;

        target = Content.ProblemData.Dataset.GetDoubleValues(Content.ProblemData.TargetVariable).ToArray();
        switch (SamplesComboBox.SelectedItem.ToString()) {
          case SamplesComboBoxAllSamples:
            indizes = Enumerable.Range(0, Content.ProblemData.Dataset.Rows).ToArray();
            estimatedClassValues = Content.EstimatedClassValues.ToArray();
            confidences = weightCalc.GetConfidence(solutions,
                                                   Enumerable.Range(0, Content.ProblemData.Dataset.Rows),
                                                   estimatedClassValues,
                                                   weightCalc.GetAllClassDelegate()).ToArray();
            break;
          case SamplesComboBoxTrainingSamples:
            indizes = Content.ProblemData.TrainingIndices.ToArray();
            estimatedClassValues = Content.EstimatedTrainingClassValues.ToArray();
            confidences = weightCalc.GetConfidence(solutions,
                                                   Content.ProblemData.TrainingIndices,
                                                   estimatedClassValues,
                                                   weightCalc.GetTrainingClassDelegate()).ToArray();
            break;
          case SamplesComboBoxTestSamples:
            indizes = Content.ProblemData.TestIndices.ToArray();
            estimatedClassValues = Content.EstimatedTestClassValues.ToArray();
            confidences = weightCalc.GetConfidence(solutions,
                                                   Content.ProblemData.TestIndices,
                                                   estimatedClassValues,
                                                   weightCalc.GetTestClassDelegate()).ToArray();
            break;
          default:
            throw new ArgumentException();
        }

        if (!estimatedClassValues.All(x => Double.IsNaN(x))) {
          int row;
          for (int i = 0; i < maxPoints; i++) {
            double confidenceValue = (1.0 / (maxPoints - 1)) * i;
            int notCovered = 0;

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

            accuracy[i + 1] = accuracyCalc.Accuracy;
            if (indizes.Length > 0) {
              covered[i] = 1.0 - (double)notCovered / (double)indizes.Length;
            }
            accuracyCalc.Reset();
          }

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

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

          Series area = this.chart.Series.Add(AREA);
          area.ChartType = SeriesChartType.Area;
          area.Color = Color.LightBlue;
          IEnumerable<IEnumerable<double>> areaPoints = CalculateAreaPoints(covered, accuracy);
          area.Points.DataBindXY(areaPoints.ElementAt(0), areaPoints.ElementAt(1));

          Series series = this.chart.Series.Add(ACCURACYCOVERED);
          series.Color = Color.Red;
          series.ChartType = SeriesChartType.FastPoint;
          series.MarkerStyle = MarkerStyle.Diamond;
          series.MarkerSize = 5;
          series.Points.DataBindXY(covered, accuracy);

          double auc = CalculateAreaUnderCurve(series);
          area.LegendToolTip = "AUC: " + auc;

          AUCLabel.Text = "AUC: " + auc;
        } else {
          AUCLabel.Text = "No values in this partition!";
        }
      }
    }

    private IEnumerable<IEnumerable<double>> CalculateAreaPoints(double[] covered, double[] accuracy) {
      List<double> newCovered = new List<double>();
      List<double> worseAccuracy = new List<double>();
      newCovered.Add(covered[0]);
      worseAccuracy.Add(accuracy[0]);
      for (int i = 1; i < covered.Length; i++) {
        if (accuracy[i] > accuracy[i - 1]) {
          worseAccuracy.Add(accuracy[i - 1]);
          newCovered.Add(covered[i] - Double.Epsilon);
        } else {
          worseAccuracy.Add(accuracy[i]);
          newCovered.Add(covered[i - 1] + Double.Epsilon);
        }
        worseAccuracy.Add(accuracy[i]);
        newCovered.Add(covered[i]);
      }
      return new List<IEnumerable<double>>() { newCovered, worseAccuracy };
    }

    private double CalculateAreaUnderCurve(Series series) {
      if (series.Points.Count < 1) throw new ArgumentException("Could not calculate area under curve if less than 1 data points were given.");

      double auc = 0.0;
      for (int i = 1; i < series.Points.Count; i++) {
        double width = series.Points[i].XValue - series.Points[i - 1].XValue;
        double y1 = series.Points[i - 1].YValues[0];
        double y2 = series.Points[i].YValues[0];

        auc += (y1 + y2) * width / 2;
      }

      return auc;
    }

    #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
  }
}