/*
* RapidMiner
*
* Copyright (C) 2001-2008 by Rapid-I and the contributors
*
* Complete list of developers available at our web site:
*
* http://rapid-i.com
*
* This program is free software: you can redistribute it and/or modify
* it under the terms of the GNU Affero General Public License as published by
* the Free Software Foundation, either version 3 of the License, or
* (at your option) any later version.
*
* This program 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 Affero General Public License for more details.
*
* You should have received a copy of the GNU Affero General Public License
* along with this program. If not, see http://www.gnu.org/licenses/.
*/
package com.rapidminer.tools.math;
import java.io.Serializable;
import java.util.Arrays;
import java.util.Iterator;
import javax.swing.JDialog;
import com.rapidminer.datatable.DataTable;
import com.rapidminer.datatable.SimpleDataTable;
import com.rapidminer.datatable.SimpleDataTableRow;
import com.rapidminer.example.Attribute;
import com.rapidminer.example.AttributeTypeException;
import com.rapidminer.example.Example;
import com.rapidminer.example.ExampleSet;
import com.rapidminer.example.Statistics;
import com.rapidminer.gui.plotter.ScatterPlotter;
import com.rapidminer.gui.plotter.SimplePlotterDialog;
import com.rapidminer.gui.viewer.ROCChartPlotter;
/**
* Helper class containing some methods for ROC plots, threshold finding and
* area under curve calculation.
*
* @author Ingo Mierswa, Martin Scholz
* @version $Id: ROCDataGenerator.java,v 1.10 2008/05/09 19:23:03 ingomierswa Exp $
*/
public class ROCDataGenerator implements Serializable {
private static final long serialVersionUID = -4473681331604071436L;
/** Defines the maximum amount of points which is plotted in the ROC curve. */
public static final int MAX_ROC_POINTS = 200;
private double misclassificationCostsPositive = 1.0d;
private double misclassificationCostsNegative = 1.0d;
private double slope = 1.0d;
private double bestThreshold = Double.NaN;
/** Creates a new ROC data generator. */
public ROCDataGenerator(double misclassificationCostsPositive, double misclassificationCostsNegative) {
this.misclassificationCostsPositive = misclassificationCostsPositive;
this.misclassificationCostsNegative = misclassificationCostsNegative;
}
/** The best threshold will automatically be determined during the calculation of the
* ROC data list. Please note that the given weights are taken into account (defining
* the slope. */
public double getBestThreshold() {
return bestThreshold;
}
/** Creates a list of ROC data points from the given example set. The example set must have
* a binary label attribute and confidence values for both values, i.e. a model must have been
* applied on the data. */
public ROCData createROCData(ExampleSet exampleSet, boolean useExampleWeights) {
Attribute label = exampleSet.getAttributes().getLabel();
exampleSet.recalculateAttributeStatistics(label);
Attribute predictedLabel = exampleSet.getAttributes().getPredictedLabel();
// create sorted collection with all label values and example weights
WeightedConfidenceAndLabel[] calArray = new WeightedConfidenceAndLabel[exampleSet.size()];
Attribute weightAttr = null;
if (useExampleWeights)
weightAttr = exampleSet.getAttributes().getWeight();
Attribute labelAttr = exampleSet.getAttributes().getLabel();
String positiveClassName = null;
if (label.isNominal() && (label.getMapping().size() == 2)) {
positiveClassName = labelAttr.getMapping().mapIndex(label.getMapping().getPositiveIndex());
} else if (label.isNominal() && (label.getMapping().size() == 1)) {
positiveClassName = labelAttr.getMapping().mapIndex(0);
} else {
throw new AttributeTypeException("Cannot calculate ROC data for non-classification labels or for labels with more than 2 classes.");
}
int index = 0;
Iterator<Example> reader = exampleSet.iterator();
while (reader.hasNext()) {
Example example = reader.next();
WeightedConfidenceAndLabel wcl;
if (weightAttr == null) {
wcl = new WeightedConfidenceAndLabel(example.getConfidence(positiveClassName), example.getValue(labelAttr), example.getValue(predictedLabel));
} else {
wcl = new WeightedConfidenceAndLabel(example.getConfidence(positiveClassName), example.getValue(labelAttr), example.getValue(predictedLabel), example.getValue(weightAttr));
}
calArray[index++] = wcl;
}
Arrays.sort(calArray);
// The slope is defined by the ratio of positive examples and the
// different misclassification costs.
// The formula for the slope is (#pos / #neg) / (costs_neg / costs_pos).
double ratio =
exampleSet.getStatistics(label, Statistics.COUNT, positiveClassName) /
exampleSet.getStatistics(label, Statistics.COUNT, label.getMapping().mapIndex(label.getMapping().getNegativeIndex()));
slope = misclassificationCostsNegative / misclassificationCostsPositive;
slope = ratio / slope;
// The task is to find the isometric that crosses the TP-axis as high as
// possible
// The TP value of the best isometric seen so far is stored in
// bestIsometricsTpValue,
// the corresponding threshold is stored in bestThreshold.
double tp = 0.0d;
double sum = 0.0d;
double bestIsometricsTpValue = 0;
bestThreshold = Double.POSITIVE_INFINITY;
double oldConfidence = 1.0d;
ROCData rocData = new ROCData();
ROCPoint last = new ROCPoint(0.0d, 0.0d, 1.0d);
rocData.addPoint(last); // add first point in ROC curve
// Iterate through the example set sorted by predictions.
// In each iteration the example with next highest confidence of being
// positive
// is added to the set of covered examples.
for (int i = 0; i < calArray.length; i++) {
WeightedConfidenceAndLabel wcl = calArray[i];
double weight = wcl.getWeight();
double fp = sum - tp;
if (wcl.getLabel() == label.getMapping().getPositiveIndex()) {
tp += weight;
} else {
// c is the value at the TP axis connecting the current point in
// ROC space
// with a line with the slope given by the user.
double c = tp - (fp * slope);
if (c > bestIsometricsTpValue) {
bestIsometricsTpValue = c;
bestThreshold = wcl.getConfidence();
}
}
double currentConfidence = wcl.getConfidence();
if (currentConfidence != oldConfidence) {
rocData.addPoint(last);
oldConfidence = currentConfidence;
}
last = new ROCPoint(fp, tp, currentConfidence);
sum += weight;
}
// Calculation for last point (upper right):
double c = tp - ((sum - tp) * slope);
if (c > bestIsometricsTpValue) {
bestThreshold = Double.NEGATIVE_INFINITY;
bestIsometricsTpValue = c;
}
rocData.addPoint(new ROCPoint(sum - tp, tp, 0.0d)); // add last point in ROC curve
// scaling for plotting
rocData.setTotalPositives(tp);
rocData.setTotalNegatives(sum - tp);
rocData.setBestIsometricsTPValue(bestIsometricsTpValue / tp);
return rocData;
}
private DataTable createDataTable(ROCData data, boolean showSlope, boolean showThresholds) {
DataTable dataTable = new SimpleDataTable("ROC Plot", new String[] { "FP/N", "TP/P", "Slope", "Threshold" });
Iterator<ROCPoint> i = data.iterator();
int pointCounter = 0;
int eachPoint = Math.max(1, (int) Math.round((double) data.getNumberOfPoints() / (double) MAX_ROC_POINTS));
while (i.hasNext()) {
ROCPoint point = i.next();
if ((pointCounter == 0) || ((pointCounter % eachPoint) == 0) || (!i.hasNext())) { // draw only MAX_ROC_POINTS points
double fpRate = point.getFalsePositives() / data.getTotalNegatives();
double tpRate = point.getTruePositives() / data.getTotalPositives();
double threshold = point.getConfidence();
dataTable.add(new SimpleDataTableRow(new double[] {
fpRate, // x
tpRate, // y1
data.getBestIsometricsTPValue() + (fpRate * slope * (data.getTotalNegatives() / data.getTotalPositives())), // y2: slope
threshold // y3: threshold or confidence
}));
}
pointCounter++;
}
return dataTable;
}
/** Creates a dialog containing a plotter for a given list of ROC data points. */
public void createROCPlotDialog(ROCData data, boolean showSlope, boolean showThresholds) {
SimplePlotterDialog plotter = new SimplePlotterDialog(createDataTable(data, showSlope, showThresholds));
plotter.setXAxis(0);
plotter.plotColumn(1, true);
if (showSlope)
plotter.plotColumn(2, true);
if (showThresholds)
plotter.plotColumn(3, true);
plotter.setDrawRange(0.0d, 1.0d, 0.0d, 1.0d);
plotter.setPointType(ScatterPlotter.LINES);
plotter.setSize(500, 500);
plotter.setLocationRelativeTo(plotter.getOwner());
plotter.setVisible(true);
}
/** Creates a dialog containing a plotter for a given list of ROC data points. */
public void createROCPlotDialog(ROCData data) {
ROCChartPlotter plotter = new ROCChartPlotter();
plotter.addROCData("ROC", data);
JDialog dialog = new JDialog();
dialog.setTitle("ROC Plot");
dialog.add(plotter);
dialog.setSize(500, 500);
dialog.setLocationRelativeTo(null);
dialog.setVisible(true);
}
/** Calculates the area under the curve for a given list of ROC data points. */
public double calculateAUC(ROCData rocData) {
// calculate AUC (area under curve)
double aucSum = 0.0d;
double[] last = null;
Iterator<ROCPoint> i = rocData.iterator();
while (i.hasNext()) {
ROCPoint point = i.next();
double fpDivN = point.getFalsePositives() / rocData.getTotalNegatives(); // false positives divided by sum of all negatives
double tpDivP = point.getTruePositives() / rocData.getTotalPositives(); // true positives divided by sum of all positives
/*
if (last != null) {
aucSum += ((tpDivP - last[1]) * (fpDivN - last[0]) / 2.0d) + (last[1] * (fpDivN - last[0]));
}
*/
if (last != null) {
aucSum += last[1] * (fpDivN - last[0]);
}
last = new double[] { fpDivN, tpDivP };
}
return aucSum;
}
}