In recent years, social networks are increasingly developed, and the social functions of various applications have strengthened the connection between users, which also brings a new data source for recommendation algorithms. This paper proposes a model named Knowledge-aware Network with Social Information(KNSI), which extracts users and items representations from the user’s item-space and social-space, and then effectively propagates it in the Knowledge Graph(KG) to obtain high-level features of the entity. In addition, the model uses an attention module to distinguish the contribution values of different neighbors in the KG and provide users with personalized recommendations. Experimental results on the public dataset show that the proposed model works significant in Recall and NDCG over several other models. Further analysis verify KNSI offers a valid explanation for items not interacted with by the user, alleviating the problem of sparse interactive data.