Social collaborative filtering using local dynamic overlapping community detection

Recommender systems play an important role in dealing with the problems caused by the great and growing amount of information, and the collaborative filtering method can propose high-quality suggestions through using other individuals' opinions. In real world and with the changing nature of individuals' preferences, recommender systems are not only responsible for fulfilling the users' interests, but also for modeling their dynamic behaviors. On the other hand, social networks provide new types of data that contribute to personalization and improvement of the performance of the recommender systems. In this paper, we propose a dynamic collaborating filtering-based social recommender system using a dynamic, local, and an overlapping community detection approach. In this study, in addition to the temporal users' rating data for items, temporal friendship relations among users in social network are also considered and a local community detection method is combined with social recommendation technique in order to improve scalability, sparsity, and cold start issues of collaborative filtering. The proposed method is compared with a number of state-of-the-art recommendation methods. The experimental results on benchmark datasets show that the proposed method outperforms the compared methods based on different evaluation metrics. It has fewer errors (at least 5%) and higher accuracy (at least 12%) than pervious methods in sparse rating matrix and dynamic environment.