How to construct a binary logistic classifier?

I have a matrix X containing features and a vector y containing the labels. How can I perform logistic regression and compute the accuracy of the regression as well as Type 1 and Type 2 errors? Furthermore, I want to have 5% of each category which are classified with the highest score.

X_train = [1 2 3; 4 5 6; 3 5 7];
y_train = [0 0 1];
X_test = [6 8 2; 6 3 4; 5 7 1];
y_test = [1 0 1];

[b,dev,stats] = glmfit(X_train, y_train,'binomial','logit');

But this does not work because of the dimension of y_train. And the evaluation of the classifier is missing as well. Can anybody help me please?