Recommendation Algorithm for Graph Convolutional Networks based on Multi-Ralational Knowledge Graph

Abstract

Classic recommendation technologies such as collaborative filtering have some challenging problems such as cold start. Because knowledge graph can enrich data information, in recent years, many scholars have applied it to recommendation systems to solve the above problems. However, Most of the methods only exploit relations and entities involved in the knowledge graph, and do not further explore the correlation information between the entities in the knowledge graph. In order to solve the above problems, recommendation algorithm for graph convolutional networks based on multi-relational knowledge graph (Multi-RKGCN) is proposed, which expands the relations and entities in knowledge graph through reflexive and self-circulating ways. At the time of aggregation, the tail entity and the corresponding relationship are combined and embedded by the knowledge graph embedding technology to enrich the semantics of the entity. Finally, the performance of AUC and F1 is verified on two publicly available datasets. The experimental results show that Multi-RKGCN method is better than KGCN method.