test_classifier.m

load ionosphere
resp = strcmp(Y,'b'); % resp = 1, if Y = 'b', or 0 if Y = 'g'
pred = X(:,3:34);
mdl = fitglm(pred,resp,'Distribution','binomial','Link','logit');
score_log = mdl.Fitted.Probability; % Probability estimates
[Xlog,Ylog,Tlog,AUClog] = perfcurve(resp,score_log,'true');
mdlSVM = fitcsvm(pred,resp,'Standardize',true);
mdlSVM = fitPosterior(mdlSVM);
[~,score_svm] = resubPredict(mdlSVM);
[Xsvm,Ysvm,Tsvm,AUCsvm] = perfcurve(resp,score_svm(:,mdlSVM.ClassNames),'true');
mdlNB = fitcnb(pred,resp);
[~,score_nb] = resubPredict(mdlNB);
[Xnb,Ynb,Tnb,AUCnb] = perfcurve(resp,score_nb(:,mdlNB.ClassNames),'true');
plot(Xlog,Ylog)
hold on
plot(Xsvm,Ysvm)
plot(Xnb,Ynb)
legend('Logistic Regression','Support Vector Machines','Naive Bayes','Location','Best')
xlabel('False positive rate'); ylabel('True positive rate');
title('ROC Curves for Logistic Regression, SVM, and Naive Bayes Classification')
hold off