classification - MATLAB semi supervised classifier -

i have set of 132 elements described 5 features. have labels these elements

features= rand(132,5); labels= [ones(132,1);2*ones(132,1)]; 

for classification problem found if transform features in following way:

b=pdist(features); c = squareform(b); [v,d] = eig(c); 

and take new features eigenvectors in v

new_features = v; 

i accuracy.

in particular train supervised classifier using new_features , test in 10 fold cross validation.

the problem not know transformation feature new_features unknown new element new_element = rand(1,5).

i thinking train semi-supervised classifier (without label of new unknown element, using transformation) , test performance on unknown (unlabelled) elements.

could recommend matlab implementation of semi-supervised classifier or way transform unknown new element in eigenvector space of known elements (used training of supervised classifier)?