knnclass.m

一个关于k最近临法分类的matlab程序

% load data files
train=load('spambase_train.txt','-ascii');
train(:,end)=train(:,end)+1;
trainLabels=train(:,end);
numTraining=length(trainLabels);

test=load('spambase_test.txt','-ascii');
test(:,end)=test(:,end)+1;
testLabels=test(:,end);
numTesting=length(testLabels);
test(:,end)=[];

uniqueClassLabels=unique(trainLabels);      % unique class labels
numClasses = length(uniqueClassLabels);     % number of classes

% show the training data and testing data
figure; 
subplot(3,2,1);
scatter(train(:,1),train(:,2),3,trainLabels,'.');
title('training data');
set(gca,'xlim',[-4 4], 'ylim', [-6,4]);

subplot(3,2,2);
scatter(test(:,1),test(:,2),3,testLabels,'+');
title('testing data');
set(gca,'xlim',[-4 4], 'ylim', [-6,4]);

k=[ 1 3 5 7];
for i=1:length(k)
    [predictLabels]=knn(train,test,k(i)); % knn classifier
    subplot(3,2,i+2);
    scatter(test(:,1),test(:,2),5,predictLabels,'+');
    hold on;
    errorIdx=predictLabels~=testLabels;
    scatter(test(errorIdx,1),test(errorIdx,2),3,'*r');
    set(gca,'xlim',[-4 4], 'ylim', [-6,4]);
    hold off;
    pError=sum(predictLabels~=testLabels)/numTesting; % if predicted label does not match the true label, we get a error

    title(['k = ' num2str(k(i)) ', errorRate = '  num2str(pError)]);


%  do we need to display the density matrix?
%     disp(['Density estimated by parzen window with window width of ', num2str(windowWidth(i)), ' is:'] );
%     disp(estDensity);
    
    disp(['Error rate for k = ', num2str(k(i)), ' is: ' num2str(pError)] ); % display the error rate.

end
suptitle('knn classifier on spambase');