Learning over No-Preferred and Preferred Sequence of Items for Robust Recommendation