智能系统学报2023，Vol.18Issue(6)：1295-1304,10.DOI:10.11992/tis.202203039
融合图卷积注意力机制的协同过滤推荐方法
Collaborative filtering recommendation approach fused with graph convolutional attention mechanism
摘要
Abstract
The graph convolutional neural network(GCN)has attracted extensive attention due to its powerful model-ing capabilities.In item recommendation,existing graph convolution collaborative filtering techniques ignore the im-portance of neighbor nodes in the propagation aggregation process,making the embedding vector representation of user and item unreasonable.Therefore,this paper proposes a collaborative filtering recommendation model fused with graph convolutional attention to address this problem.First,user-item interaction information was mapped to a low-dimension-al,dense vector space using graph embedding techniques.Further,the high-order interaction information between the user and the item was learned using stacking multiple layers of GCN.The model also fused attention mechanisms to ad-aptively assign weights to neighbor nodes,thereby capturing the influence of highly representative neighbors.Simultan-eously,the model could rely only on feature expressions between nodes when aggregating feature information from neighboring nodes,increasing the independence of the graph structure and improving the generalization capability of the model.Finally,a hierarchical aggregation function that aggregated multiple embedding vectors,which was learned from the graph convolution layer by weighting,was designed,and the inner product function was used to obtain the associ-ation score between the user and the item.Results of the extensive experiments conducted on three real datasets have demonstrated the effectiveness of the proposed approach.关键词
图嵌入技术/图卷积神经网络/注意力机制/协同过滤/用户偏好/协同过滤/高阶交互/邻域聚合Key words
graph embedding technology/graph convolutional network/attention mechanism/collaborative filtering/user preference/collaborative filtering/high-order interaction/neighbor aggregation分类
计算机与自动化引用本文复制引用
朱金侠,孟祥福,邢长征,张霄雁..融合图卷积注意力机制的协同过滤推荐方法[J].智能系统学报,2023,18(6):1295-1304,10.基金项目
国家重点研发计划项目(2018YFB1402901) （2018YFB1402901）
国家自然科学基金项目(61772249) （61772249）
辽宁省教育厅一般项目(LJ2019QL017). （LJ2019QL017）
