区分交互意图的图卷积协同过滤算法

Graph convolution collaborative filtering algorithm for discriminating interaction intentions

近几年提出了一些基于图卷积网络的协同过滤推荐模型,然而大部分模型将邻域权重视为常量且不区分用户和物品间的交互关系,无法获取令用户满意的推荐列表。因此,为了得到用户和物品更准确的嵌入表示,提出一种区分交互意图的图卷积协同过滤推荐算法MiGCCF(multi-intention graph convolutional collaborative filtering)。该算法将交互关系进行分解,细粒度分析用户与物品间的交互意图,并引入注意力机制,在消息传播过程中赋予邻域可学习的注意力权重,挖掘用户对于不同交互物品的喜爱度。在Gowalla与Amazon-book上的实验表明,该算法相比于基准算法,在两个数据集上的HR@50和NDCG@50指标分别提高了12.5%和8.5%,具有更好的性能表现。

In recent years,some collaborative filtering recommendation models based on graph convolutional networks have been proposed.However,most models regard neighborhood weights as constants and do not distinguish the interaction between users and items,so they cannot obtain a satisfactory recommendation list for users.In order to get a more accurate embedded representation of users and items,this paper proposed a multi-intention-based graph convolution collaborative filtering recommendation algorithm MiGCCF.The algorithm decomposed the interaction relationship,it analyzed the interaction intention between the user and the item in a fine-grained manner,and introduced an attention mechanism to give the neighborhood a learnable attention weight in the process of message dissemination,and to mine the user’s preference for different interactive items.Experiments on Gowalla and Amazon-book show that compared with the benchmark algorithm,the HR@50 and NDCG@50 indicators of the two datasets are improved by 12.5%and 8.5%,respectively,with better performance.