How MATLAB classifier learner app divides the feature set into training , validation and testing data?

Does this app itself divides the feature set in training ,validation and test data?

actually m getting an accuracy of 99% every time, which is impossible on my dataset.

Used an alternative approach , where i myself divide the feature set in training and testing. First run the classifier on training then extract the model and finally fed the left features to classifier.

without labels, Still getting same accuracy?

Cud somebody guide?