prtDecisionBinaryMinPe Decision object for minimum probability of
error in binary classification
prtDec = prtDecisionBinaryMinPe creates a prtDecisionBinaryMinPe
object, which can be used find a decision threshold in a binary
classification problem that minimizes the probability of error.
prtDecision objects are intended to be used either as members of
prtAlgorithm or prtClass objects.
Example 1:
ds = prtDataGenBimodal; % Load a data set
classifier = prtClassKnn; % Create a clasifier
classifier = classifier.train(ds); % Train the classifier
yOutClassifier = classifier.run(ds); % Run the classifier
% Construct a prtAlgorithm object consisting of a prtClass object and
% a prtDecision object
algo = prtClassKnn + prtDecisionBinaryMinPe;
algo = algo.train(ds); % Train the algorithm
yOutAlgorithm = algo.run(ds); % Run the algorithm
% Plot and compare the results
subplot(2,1,1); stem(yOutClassifier.getObservations); title('KNN Output');
subplot(2,1,2); stem(yOutAlgorithm.getObservations); title('KNN + Decision Output');
Example 2:
ds = prtDataGenBimodal; % Load a data set
classifier = prtClassKnn; % Create a clasifier
classifier = classifier.train(ds); % Train the classifier
% Plot the trained classifier
subplot(2,1,1); plot(classifier); title('KNN');
% Set the classifiers internealDecider to be a prtDecsion object
classifier.internalDecider = prtDecisionBinaryMinPe;
classifier = classifier.train(ds); % Train the classifier
subplot(2,1,2); plot(classifier); title('KNN + Decision');