svm matlab 画图,关于svm 画图问题

load wine_SVM;

data = [wine(:,1), wine(:,2)];

groups = ismember(wine_labels,1); % not familiar with ismember? help... help...

[train, test] = crossvalind('holdOut',groups); % not familiar with crossvalind? help... help...

%train and test here are logical format

train_wine = data(train,:);

train_wine_labels = groups(train,:);

train_wine_labels = double( train_wine_labels );

%in the usage of libsvm, the label vector and instance matrix must be

%double. So we have to transfer the format of labels in this case.

test_wine = data(test,:);

test_wine_labels = groups(test,:);

test_wine_labels = double( test_wine_labels );

% train_wine = normalization(train_wine',2);

% test_wine = normalization(test_wine',2);

% train_wine = train_wine';

% test_wine = test_wine';

% bestcv = 0;

% for log2c = -5:5,

%   for log2g = -5:5,

%     cmd = ['-v 5 -c ', num2str(2^log2c), ' -g ', num2str(2^log2g)];

%     cv = svmtrain(train_wine_labels, train_wine, cmd);

%     if (cv >= bestcv),

%       bestcv = cv; bestc = 2^log2c; bestg = 2^log2g;

%     end

%   end

% end

% fprintf('%g %g %g (best c=%g, g=%g, rate=%g)\n', log2c, log2g, cv, bestc, bestg, bestcv);

% cmd = ['-c ', num2str(bestc), ' -g ', num2str(bestg)];

% model = svmtrain(train_wine_labels, train_wine, cmd);

model = svmtrain(train_wine_labels, train_wine, '-c 1 -g 0.07');

[predict_label, accuracy] = svmpredict(test_wine_labels, test_wine, model);

%plot visualization for train_wine

% w = model.SVs' * model.sv_coef;

% b = -model.rho;

[mm,mn] = size(model.SVs);

figure;

hold on;

[m,n] = size(train_wine);

for run = 1:m

if train_wine_labels(run) == 1

h1 = plot( train_wine(run,1),train_wine(run,2),'r+' );

else

h2 = plot( train_wine(run,1),train_wine(run,2),'g*' );

end

for i = 1:mm

if model.SVs(i,1)==train_wine(run,1) && model.SVs(i,2)==train_wine(run,2)

h3 = plot( train_wine(run,1),train_wine(run,2),'o' );

end

end

end

legend([h1,h2,h3],'1','0','Support Vectors');

hold off;