Scalable User Similarity Estimation Based on fuzzy proximity for Enhancing Accuracy of Collaborative Filtering Recommendation

User similarity measurement plays a key role in collaborative filtering based recommender systems. In order to improve accuracy and scalability of traditional user based collaborative filtering techniques under the conditions of largescale and spars data, we make some contributions. We define two indices of Homophily Correlation and Influence Correlation from the most popular social phenomena which include user interest synthesize and accordingly, we describe a proximity based similarity measurement model using fuzzy inference system. Finally, we demonstrate effectiveness of the proposed similarity measure in recommended performance on a real movie rating data set, the MovieLens data set, compared to state-of-theart methods.