Social Recommendation Based On Deep User Interest Modeling

Collaborative Filtering(CF)is one of the most popular techniques for building recommender systems.To overcome the data sparsity in CF,social recommender systems have emerged to boost recommendation performance by utilizing social correlation among users' interests.Recently,inspired by the immense success of deep learning for embedding learning,neural network based recommender systems have shown promising recommendation performance.There exist two research tasks in our work,static social recommendation and temporal social recommendation.For static social recommendation,there exists two key challenges:First,how to incorporate the social correlation of users'interests?Second,how to design a neural architecture to capture the unique characteristics of user-item interaction behavior for recommendation?For temporal social recommendation,there also exists two key challenges:Third,how to capture users'dynamic and static interests?Fourth,how to model the complex interplay between users'interests and social influence over time?To tackle the challenges in static social recommendation,we develop a model named CNSR(Collaborative Neural Social Recommendation)with two parts:a social embedding part and a collaborative neural recommendation part.In CNSR,the user embedding leverages each user's social embedding learned from an unsupervised deep learning technique with social correlation regularization.The user and item embedding are then fed into a unique neural network with a newly designed collaboration layer to model both the shallow collaborative and deep complex interaction relationships between users and items.We further propose a joint learning framework to allow the social embedding part and the collaborative neural recommendation part to mutually enhance each other.To address these challenges in temporal social recommendation,we present an attentive recurrent network based approach for temporal social recommendation.In the proposed approach,we model users' complex dynamic and general static preference over time by leveraging social influence among users with two attention networks.Specifically,in the dynamic preference modeling process,we design a temporal attention network to model the temporal social influence over time,and a dynamic social aware recurrent neural network is proposed to capture users'complex latent interests over time.In the general static preference modeling process,we augment each user's static interest part by introducing a static social attention module to model the stationary social influence among users.The output of the dynamic preferences and the static preferences are fused together in a unified framework for the temporal social recommendation task.In this paper,in order to tackle the two tasks,we design a neural architecture that organically combines the intrinsic relationship between social network structure and user-item interaction behavior for static recommendation.And we present an attentive recurrent network based approach for temporal social recommendation.Finally,experimental results on two real-world datasets show the superiority of our proposed model compared to the baselines.