[4417] | 1 | #region License Information
|
---|
| 2 | /* HeuristicLab
|
---|
[17180] | 3 | * Copyright (C) Heuristic and Evolutionary Algorithms Laboratory (HEAL)
|
---|
[4417] | 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.Drawing;
|
---|
| 25 | using System.Linq;
|
---|
| 26 | using System.Text;
|
---|
| 27 | using System.Windows.Forms;
|
---|
| 28 | using System.Windows.Forms.DataVisualization.Charting;
|
---|
| 29 | using HeuristicLab.Common;
|
---|
| 30 | using HeuristicLab.MainForm;
|
---|
| 31 | using HeuristicLab.MainForm.WindowsForms;
|
---|
[5829] | 32 | namespace HeuristicLab.Problems.DataAnalysis.Views {
|
---|
[5975] | 33 | [View("ROC Curves")]
|
---|
[5664] | 34 | [Content(typeof(IDiscriminantFunctionClassificationSolution))]
|
---|
[6642] | 35 | public partial class DiscriminantFunctionClassificationRocCurvesView : DataAnalysisSolutionEvaluationView {
|
---|
[4417] | 36 | private const string xAxisTitle = "False Positive Rate";
|
---|
| 37 | private const string yAxisTitle = "True Positive Rate";
|
---|
| 38 | private const string TrainingSamples = "Training";
|
---|
| 39 | private const string TestSamples = "Test";
|
---|
| 40 | private Dictionary<string, List<ROCPoint>> cachedRocPoints;
|
---|
| 41 |
|
---|
[5664] | 42 | public DiscriminantFunctionClassificationRocCurvesView() {
|
---|
[4417] | 43 | InitializeComponent();
|
---|
| 44 |
|
---|
| 45 | cachedRocPoints = new Dictionary<string, List<ROCPoint>>();
|
---|
| 46 |
|
---|
| 47 | cmbSamples.Items.Add(TrainingSamples);
|
---|
| 48 | cmbSamples.Items.Add(TestSamples);
|
---|
| 49 | cmbSamples.SelectedIndex = 0;
|
---|
| 50 |
|
---|
[4651] | 51 | chart.CustomizeAllChartAreas();
|
---|
[4417] | 52 | chart.ChartAreas[0].AxisX.Minimum = 0.0;
|
---|
| 53 | chart.ChartAreas[0].AxisX.Maximum = 1.0;
|
---|
| 54 | chart.ChartAreas[0].AxisX.MajorGrid.Interval = 0.2;
|
---|
| 55 | chart.ChartAreas[0].AxisY.Minimum = 0.0;
|
---|
| 56 | chart.ChartAreas[0].AxisY.Maximum = 1.0;
|
---|
| 57 | chart.ChartAreas[0].AxisY.MajorGrid.Interval = 0.2;
|
---|
| 58 |
|
---|
| 59 | chart.ChartAreas[0].AxisX.Title = xAxisTitle;
|
---|
| 60 | chart.ChartAreas[0].AxisY.Title = yAxisTitle;
|
---|
| 61 | }
|
---|
| 62 |
|
---|
[5664] | 63 | public new IDiscriminantFunctionClassificationSolution Content {
|
---|
| 64 | get { return (IDiscriminantFunctionClassificationSolution)base.Content; }
|
---|
[4417] | 65 | set { base.Content = value; }
|
---|
| 66 | }
|
---|
| 67 |
|
---|
| 68 | protected override void RegisterContentEvents() {
|
---|
| 69 | base.RegisterContentEvents();
|
---|
[5664] | 70 | Content.ModelChanged += new EventHandler(Content_ModelChanged);
|
---|
[4417] | 71 | Content.ProblemDataChanged += new EventHandler(Content_ProblemDataChanged);
|
---|
| 72 | }
|
---|
| 73 | protected override void DeregisterContentEvents() {
|
---|
| 74 | base.DeregisterContentEvents();
|
---|
[5664] | 75 | Content.ModelChanged -= new EventHandler(Content_ModelChanged);
|
---|
[4417] | 76 | Content.ProblemDataChanged -= new EventHandler(Content_ProblemDataChanged);
|
---|
| 77 | }
|
---|
| 78 |
|
---|
[5664] | 79 | private void Content_ModelChanged(object sender, EventArgs e) {
|
---|
[4417] | 80 | UpdateChart();
|
---|
| 81 | }
|
---|
| 82 | private void Content_ProblemDataChanged(object sender, EventArgs e) {
|
---|
| 83 | UpdateChart();
|
---|
| 84 | }
|
---|
| 85 |
|
---|
| 86 | protected override void OnContentChanged() {
|
---|
| 87 | base.OnContentChanged();
|
---|
| 88 | chart.Series.Clear();
|
---|
| 89 | if (Content != null) UpdateChart();
|
---|
| 90 | }
|
---|
| 91 |
|
---|
| 92 | private void UpdateChart() {
|
---|
| 93 | if (InvokeRequired) Invoke((Action)UpdateChart);
|
---|
| 94 | else {
|
---|
| 95 | chart.Series.Clear();
|
---|
| 96 | chart.Annotations.Clear();
|
---|
| 97 | cachedRocPoints.Clear();
|
---|
| 98 |
|
---|
| 99 | int slices = 100;
|
---|
[4469] | 100 | IEnumerable<int> rows;
|
---|
[4417] | 101 |
|
---|
| 102 | if (cmbSamples.SelectedItem.ToString() == TrainingSamples) {
|
---|
[8139] | 103 | rows = Content.ProblemData.TrainingIndices;
|
---|
[4417] | 104 | } else if (cmbSamples.SelectedItem.ToString() == TestSamples) {
|
---|
[8139] | 105 | rows = Content.ProblemData.TestIndices;
|
---|
[4417] | 106 | } else throw new InvalidOperationException();
|
---|
| 107 |
|
---|
[4469] | 108 | double[] estimatedValues = Content.GetEstimatedValues(rows).ToArray();
|
---|
[6740] | 109 | double[] targetClassValues = Content.ProblemData.Dataset.GetDoubleValues(Content.ProblemData.TargetVariable, rows).ToArray();
|
---|
[4417] | 110 | double minThreshold = estimatedValues.Min();
|
---|
| 111 | double maxThreshold = estimatedValues.Max();
|
---|
| 112 | double thresholdIncrement = (maxThreshold - minThreshold) / slices;
|
---|
| 113 | minThreshold -= thresholdIncrement;
|
---|
| 114 | maxThreshold += thresholdIncrement;
|
---|
| 115 |
|
---|
[8831] | 116 | List<double> classValues = Content.ProblemData.ClassValues.ToList();
|
---|
[4417] | 117 |
|
---|
| 118 | foreach (double classValue in classValues) {
|
---|
| 119 | List<ROCPoint> rocPoints = new List<ROCPoint>();
|
---|
| 120 | int positives = targetClassValues.Where(c => c.IsAlmost(classValue)).Count();
|
---|
[4469] | 121 | int negatives = targetClassValues.Length - positives;
|
---|
[4417] | 122 |
|
---|
| 123 | for (double lowerThreshold = minThreshold; lowerThreshold < maxThreshold; lowerThreshold += thresholdIncrement) {
|
---|
| 124 | for (double upperThreshold = lowerThreshold + thresholdIncrement; upperThreshold < maxThreshold; upperThreshold += thresholdIncrement) {
|
---|
[5417] | 125 | //only adapt lower threshold for binary classification problems and upper class prediction
|
---|
| 126 | if (classValues.Count == 2 && classValue == classValues[1]) upperThreshold = double.PositiveInfinity;
|
---|
| 127 |
|
---|
[4417] | 128 | int truePositives = 0;
|
---|
| 129 | int falsePositives = 0;
|
---|
| 130 |
|
---|
| 131 | for (int row = 0; row < estimatedValues.Length; row++) {
|
---|
| 132 | if (lowerThreshold < estimatedValues[row] && estimatedValues[row] < upperThreshold) {
|
---|
| 133 | if (targetClassValues[row].IsAlmost(classValue)) truePositives++;
|
---|
| 134 | else falsePositives++;
|
---|
| 135 | }
|
---|
| 136 | }
|
---|
| 137 |
|
---|
| 138 | double truePositiveRate = ((double)truePositives) / positives;
|
---|
| 139 | double falsePositiveRate = ((double)falsePositives) / negatives;
|
---|
| 140 |
|
---|
| 141 | ROCPoint rocPoint = new ROCPoint(truePositiveRate, falsePositiveRate, lowerThreshold, upperThreshold);
|
---|
[6912] | 142 | if (!rocPoints.Any(x => x.TruePositiveRate >= rocPoint.TruePositiveRate && x.FalsePositiveRate <= rocPoint.FalsePositiveRate)) {
|
---|
| 143 | rocPoints.RemoveAll(x => x.FalsePositiveRate >= rocPoint.FalsePositiveRate && x.TruePositiveRate <= rocPoint.TruePositiveRate);
|
---|
[4417] | 144 | rocPoints.Add(rocPoint);
|
---|
| 145 | }
|
---|
| 146 | }
|
---|
[5417] | 147 | //only adapt upper threshold for binary classification problems and upper class prediction
|
---|
| 148 | if (classValues.Count == 2 && classValue == classValues[0]) lowerThreshold = double.PositiveInfinity;
|
---|
[4417] | 149 | }
|
---|
| 150 |
|
---|
| 151 | string className = Content.ProblemData.ClassNames.ElementAt(classValues.IndexOf(classValue));
|
---|
[6912] | 152 | cachedRocPoints[className] = rocPoints.OrderBy(x => x.FalsePositiveRate).ToList(); ;
|
---|
[4417] | 153 |
|
---|
| 154 | Series series = new Series(className);
|
---|
| 155 | series.ChartType = SeriesChartType.Line;
|
---|
| 156 | series.MarkerStyle = MarkerStyle.Diamond;
|
---|
| 157 | series.MarkerSize = 5;
|
---|
| 158 | chart.Series.Add(series);
|
---|
| 159 | FillSeriesWithDataPoints(series, cachedRocPoints[className]);
|
---|
| 160 |
|
---|
| 161 | double auc = CalculateAreaUnderCurve(series);
|
---|
| 162 | series.LegendToolTip = "AUC: " + auc;
|
---|
| 163 | }
|
---|
| 164 | }
|
---|
| 165 | }
|
---|
| 166 |
|
---|
| 167 | private void FillSeriesWithDataPoints(Series series, IEnumerable<ROCPoint> rocPoints) {
|
---|
| 168 | series.Points.Add(new DataPoint(0, 0));
|
---|
| 169 | foreach (ROCPoint rocPoint in rocPoints) {
|
---|
| 170 | DataPoint point = new DataPoint();
|
---|
[6912] | 171 | point.XValue = rocPoint.FalsePositiveRate;
|
---|
| 172 | point.YValues[0] = rocPoint.TruePositiveRate;
|
---|
[4417] | 173 | point.Tag = rocPoint;
|
---|
| 174 |
|
---|
| 175 | StringBuilder sb = new StringBuilder();
|
---|
[6912] | 176 | sb.AppendLine("True Positive Rate: " + rocPoint.TruePositiveRate);
|
---|
| 177 | sb.AppendLine("False Positive Rate: " + rocPoint.FalsePositiveRate);
|
---|
| 178 | sb.AppendLine("Upper Threshold: " + rocPoint.UpperThreshold);
|
---|
| 179 | sb.AppendLine("Lower Threshold: " + rocPoint.LowerThreshold);
|
---|
[4417] | 180 | point.ToolTip = sb.ToString();
|
---|
| 181 |
|
---|
| 182 | series.Points.Add(point);
|
---|
| 183 | }
|
---|
| 184 | series.Points.Add(new DataPoint(1, 1));
|
---|
| 185 | }
|
---|
| 186 |
|
---|
| 187 | private double CalculateAreaUnderCurve(Series series) {
|
---|
| 188 | if (series.Points.Count < 1) throw new ArgumentException("Could not calculate area under curve if less than 1 data points were given.");
|
---|
| 189 |
|
---|
| 190 | double auc = 0.0;
|
---|
| 191 | for (int i = 1; i < series.Points.Count; i++) {
|
---|
| 192 | double width = series.Points[i].XValue - series.Points[i - 1].XValue;
|
---|
| 193 | double y1 = series.Points[i - 1].YValues[0];
|
---|
| 194 | double y2 = series.Points[i].YValues[0];
|
---|
| 195 |
|
---|
| 196 | auc += (y1 + y2) * width / 2;
|
---|
| 197 | }
|
---|
| 198 |
|
---|
| 199 | return auc;
|
---|
| 200 | }
|
---|
| 201 |
|
---|
| 202 | private void cmbSamples_SelectedIndexChanged(object sender, System.EventArgs e) {
|
---|
| 203 | if (Content != null)
|
---|
| 204 | UpdateChart();
|
---|
| 205 | }
|
---|
| 206 |
|
---|
| 207 |
|
---|
| 208 | #region show / hide series
|
---|
| 209 | private void ToggleSeries(Series series) {
|
---|
| 210 | if (series.Points.Count == 0)
|
---|
| 211 | FillSeriesWithDataPoints(series, cachedRocPoints[series.Name]);
|
---|
| 212 | else
|
---|
| 213 | series.Points.Clear();
|
---|
| 214 | }
|
---|
| 215 | private void chart_MouseDown(object sender, MouseEventArgs e) {
|
---|
| 216 | HitTestResult result = chart.HitTest(e.X, e.Y);
|
---|
| 217 | if (result.ChartElementType == ChartElementType.LegendItem) {
|
---|
| 218 | if (result.Series != null) ToggleSeries(result.Series);
|
---|
| 219 | }
|
---|
| 220 | }
|
---|
| 221 | private void chart_CustomizeLegend(object sender, CustomizeLegendEventArgs e) {
|
---|
| 222 | foreach (LegendItem legendItem in e.LegendItems) {
|
---|
| 223 | var series = chart.Series[legendItem.SeriesName];
|
---|
| 224 | if (series != null) {
|
---|
| 225 | bool seriesIsInvisible = series.Points.Count == 0;
|
---|
| 226 | foreach (LegendCell cell in legendItem.Cells)
|
---|
| 227 | cell.ForeColor = seriesIsInvisible ? Color.Gray : Color.Black;
|
---|
| 228 | }
|
---|
| 229 | }
|
---|
| 230 | }
|
---|
| 231 | private void chart_MouseMove(object sender, MouseEventArgs e) {
|
---|
| 232 | HitTestResult result = chart.HitTest(e.X, e.Y);
|
---|
| 233 | if (result.ChartElementType == ChartElementType.LegendItem)
|
---|
| 234 | this.Cursor = Cursors.Hand;
|
---|
| 235 | else
|
---|
| 236 | this.Cursor = Cursors.Default;
|
---|
| 237 |
|
---|
| 238 | string newTooltipText = string.Empty;
|
---|
| 239 | if (result.ChartElementType == ChartElementType.DataPoint)
|
---|
| 240 | newTooltipText = ((DataPoint)result.Object).ToolTip;
|
---|
| 241 |
|
---|
| 242 | string oldTooltipText = this.toolTip.GetToolTip(chart);
|
---|
| 243 | if (newTooltipText != oldTooltipText)
|
---|
| 244 | this.toolTip.SetToolTip(chart, newTooltipText);
|
---|
| 245 | }
|
---|
| 246 | #endregion
|
---|
| 247 |
|
---|
| 248 |
|
---|
| 249 | private class ROCPoint {
|
---|
| 250 | public ROCPoint(double truePositiveRate, double falsePositiveRate, double lowerThreshold, double upperThreshold) {
|
---|
[6912] | 251 | this.TruePositiveRate = truePositiveRate;
|
---|
| 252 | this.FalsePositiveRate = falsePositiveRate;
|
---|
| 253 | this.LowerThreshold = lowerThreshold;
|
---|
| 254 | this.UpperThreshold = upperThreshold;
|
---|
[4417] | 255 |
|
---|
| 256 | }
|
---|
[6912] | 257 | public double TruePositiveRate { get; private set; }
|
---|
| 258 | public double FalsePositiveRate { get; private set; }
|
---|
| 259 | public double LowerThreshold { get; private set; }
|
---|
| 260 | public double UpperThreshold { get; private set; }
|
---|
[4417] | 261 | }
|
---|
| 262 |
|
---|
| 263 | }
|
---|
| 264 | }
|
---|