parameterTuning.m

function accuracy = parameterTuning( kernel_function, sigma, c )
%UNTITLED Summary of this function goes here
%   Detailed explanation goes here
net = svm(size(normal_features, 2), kernel_function, [sigma], c);
pos_class = ones(size(normal_train_sunset, 1), 1);
neg_class = ones(size(normal_train_nonsunset, 1), 1);
neg_class(:) = -1;
yTrain = vertcat(pos_class, neg_class);
xTrain = vertcat(normal_train_sunset, normal_train_nonsunset);
net = svmtrain(net, xTrain, yTrain);

falsePos = 0;
falseNeg = 0;
truePos = 0;
trueNeg = 0;

xTest = vertcat(normal_test_sunset, normal_test_nonsunset);
[detectedClasses, distances] = svmfwd(net, xTest);
pos_test_class = ones(size(normal_test_sunset, 1), 1);
neg_test_class = ones(size(normal_test_nonsunset, 1), 1);
neg_test_class(:) = -1;
yTest = vertcat(pos_test_class, neg_test_class);

for i = 1:length(yTest)
%     fprintf('Point %d, True class: %d, detected class: %d\n', i, yTest(i), detectedClasses(i))
    if (yTest(i) == 1)
        if (detectedClasses(i) == 1)
            truePos = truePos + 1;
        else
            falseNeg = falseNeg + 1;
        end
    else
        if (detectedClasses(i) == 1)
            falsePos = falsePos + 1;
        else
            trueNeg = trueNeg + 1;
        end
    end
end

accuracy = (trueNeg + truePos)/length(yTest);
tpr = truePos/(truePos+falseNeg);
precision = truePos/(truePos+falsePos);
fpr = falsePos/(falsePos+trueNeg);

end