基于注意力机制和用户属性的图卷积网络推荐模型  

GCN recommendation model based on attention mechanism and user attributes

作　　者：张荣梅[1] 李甜甜 张佳惠 ZHANG Rongmei;LI Tiantian;ZHANG Jiahui(College of Information Technology,Hebei University of Economics and Business,Shijiazhuang 050061,China)

机构地区：[1]河北经贸大学信息技术学院,河北石家庄050061

出　　处：《传感器与微系统》2024年第5期129-132,共4页Transducer and Microsystem Technologies

摘　　要：为进一步提高图卷积网络(GCN)的推荐精度和模型的收敛速度,提出了基于注意力机制和用户属性的GCN推荐模型。该模型通过轻量级GCN学习用户和项目的高阶关联信息;然后,利用注意力机制对不同邻域特征嵌入加权求和得到用户、项目潜在特征向量,利用多层感知机提取的用户属性特征向量融合到用户潜在特征向量中;最后,用户、项目潜在特征向量的内积作为预测结果进行推荐。通过在Movielens-1M数据集上实验验证,结果表明:该模型的推荐效果均优于基线模型。To further improve the recommended precision of graph convolutional network(GCN)and the convergence speed of the model,a GCN recommendation model based on attention mechanism and user attributes is proposed.The model uses a lightweight GCN to learn the high-order association information of users and items.Then,weighted summation is embedded to different neighborhood features by attention mechanism to get user and item potential feature vectors,and user attribute feature vectors extracted by multi-layer perceptron machine is fused into user potential feature vectors.Finally,the inner product of user and item potential feature vectors is recommended as a prediction result.Experimental verification on the Movielens-1M dataset show that the proposed model is prior to the baseline model.

关 键 词：推荐算法 图卷积网络 用户属性 注意力机制 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]