% exampleclassif.m April 26, 2006 R.F. Murphy
% illustrate training and testing a classifier and displaying decision boundary
%
% first generate some training and test data
% (this section could be replaced by reading some real data)
train=rand(40,2);
test=rand(10,2);
% shift the second half of the training and test data for the second variable
train(21:40,2)=train(21:40,2)+1;
test(6:10,2)=test(6:10,2)+1;
% plot the training data for the first half as blue circles and the
% second half as green diamonds
plot(train(1:20,1),train(1:20,2),'bo',train(21:40,1),train(21:40,2),'gd');
hold on
% plot the test data for the first half as blue x's and the second half
% as green plusses
plot(test(1:5,1),test(1:5,2),'bx',test(6:10,1),test(6:10,2),'g+');
hold off
% generate "labels" for each row showing that the first half is in class 1
group=[repmat(1,20,1); repmat(2,20,1)];
%
% train a linear classifier with the training data and evaluate it with
% the test data
% (this line could be replaced with a call to any other type of classifier)
cl=classify(test,train,group);
% the resulting vector shows which class the classifier thinks each test
% point is in
cl
% count true +, false +, true -, false - and build confusion matrix
tp=length(find(cl(1:5)==1));
fp=length(find(cl(6:10)==1));
fn=length(find(cl(1:5)==2));
tn=length(find(cl(6:10)==2));
confmat=[tp fn; fp tn]
%
ninc=200;
[pltx,plty]=finddecisionboundary(train(1:20,:),train(21:40,:),1,2,ninc)
% plot all of the points that were on the decision boundary
hold on
plot(pltx,plty,'k-');
hold off