prtScoreRocNfa Generate a operator characteristic curve and output the number of false alarms
prtScoreRocNfa(DECSTATS,LABELS) plots the receiver operator
characteristic curve for the decision statistics DECSTATS and the
corresponding labels LABELS. DECSTATS must be a Nx1 vector of decision
statistics. LABELS must be a Nx1 vector of binary class labels. This
behavior is the same as prtScoreRoc.
[NF,PD,THRESHOLDS,AUC] = prtScoreRocNfa(DECSTATS,LABELS) outputs the
number of false alarms NF, the probability of detection PD, the
thresholds used to obtain each PD and NF pair, and the area under the
ROC scurve AUC.
TestDataSet = prtDataGenSpiral; % Create some test and
TrainingDataSet = prtDataGenSpiral; % training data
classifier = prtClassSvm; % Create a classifier
classifier = classifier.train(TrainingDataSet); % Train
classified = run(classifier, TestDataSet);
% Find the number of false alarms at the corresponding PD.
[nf, pd ]= prtScoreRocNfa(classified.getX, TestDataSet.getY);