Research On Recommendation Algorithm With Knowledge Graph

The widespread use of the Internet has brought massive amounts of data to users.In order to reduce the impact of information overload,researchers have proposed a recommender system to make personalized recommendations for users.In recent years,the research on the combination of knowledge graph and recommender system has attracted the attention of many scholars.By studying the relationship between entities in the knowledge graph,the latent semantics of users and items can be discovered,which is helpful to improve the performance of the recommender system.However,most of the existing recommendation systems based on knowledge graphs fail to consider user preference features and item neighborhood features,and fail to effectively fuse user and item features with knowledge graphs at the same time.(1)Aiming at the problem that most recommendation models based on knowledge graphs fail to model user features effectively and do not consider the neighborhood information between entities in knowledge graphs,this paper proposes a hybrid recommendation model of knowledge graph and graph convolutional network.The model calculates the neighborhood feature vector of the item through the KGCN algorithm,and models the neighborhood relationship between entities in the knowledge graph to better utilize the structural information of the knowledge graph.At the same time,cooperative propagation and interactive unit operations are used to effectively promote Information fusion between users and entities in knowledge graphs.The model calculates the user’s features using the training method of alternating learning.Comparing experiments with seven baseline models are conducted on three public datasets,and the experimental results show that the hybrid recommendation model proposed in this paper has good recommendation performance compared with other representative recommendation models.(2)In addition,most recommendation models based on knowledge graphs fail to encode the collaborative signals in user-item interactions and do not take into account the user’s preference characteristics.This paper proposes a collaborative knowledge-aware attentive model integrating preference propagation.The model explicitly encodes collaboration signals and knowledge association signals using a heterogeneous propagation strategy,and extracts users’ preference features through a preference propagation strategy.The model in this paper is analyzed on three real datasets,and a baseline model comparison experiment,ablation experiment,sensitivity analysis experiment of parameters,and experiments on sparsity issues are designed.The experimental results show that the proposed model has good recommendation performance compared with other baseline models,and can effectively alleviate the sparsity problem.