Abstract: Collaborative filtering is one of the prevalent successful approaches in the Recommender systems to predicate items to users based on rating matrix and mitigate the difficulty of finding interesting things on the spider’s web. In this paper, we percent a Naïve Bayes model by taking into account the similarity in preferences (homophily) among the users and attributes of users (demographic) as a prior knowledge to enhance the prediction accuracy of collaborative filtering. Experiments are implemented on Movielens datasets 100K and 1M. The results show that the system can provide a recommendation in a best manner.