SBRNE: An Improved Unified Framework for Social and Behavior Recommendations with Network Embedding

Recent years have witnessed the fast growing and ubiquity of social media which has significantly changed the social manner and information sharing in our daily life. Given a user, social (i.e. friend) recommendation and behavior (i.e. item) recommendation are two types of popular services in social media applications. Despite the extensive studies, few existing work has addressed both tasks elegantly and effectively. In this paper, we propose an improved unified framework for Social and Behavior Recommendations with Network Embedding (SBRNE for short). With modeling social and behavior information simultaneously, SBRNE integrates social recommendation and behavior recommendation into a unified framework. By employing users' latent interests as a bridge, social and behavior information is modeled effectively to improve performance of social and behavior recommendations all together. In addition, an efficient network embedding procedure is introduced as a pre-training step for users' latent representations to improve effectiveness and efficiency of recommendation tasks. Extensive experiments on real-world datasets demonstrate the effectiveness of SBRNE.