Recommendation systems are widely used for information filtering designed to suggest the relevant items to the user based on user-s interest. It is necessary to predict the preferences based on the user-s historical information perfectly. Deep learning models are suitable to predict the preference knowledge by extracting the features from the user-s text and reviews about the items. Recommendations from same existing and long-term user-s interest to the users lead to a decrease in their satisfaction. However, user expectation for the recommendation service includes new trend and different range of items suggestions. Novelty and diversity are the key qualities and owns increasing attention in recommendation system. In the recommendation system, Novelty refers to the different states of suggestions with respect to previous preferences and diversity denotes the distinct set of items with respect to each other. Novelty and diversity-based recommendation systems possess the high capacity to produce an optimized recommendation list for the top-N task and improve user satisfaction. Novelty and diversity-aware top-N recommendation system with deep reference prediction yields an effective set of suggestions.