基于多层次多视角的图注意力Top-N推荐方法  

Top-N Recommendation Method for Graph Attention Based on Multi-level and Multi-view

作　　者：刘志鑫 张泽华 张杰 LIU Zhi-xin;ZHANG Ze-hua;ZHANG Jie(School of Information and Computer,Taiyuan University of Technology,Taiyuan 030024,China)

机构地区：[1]太原理工大学信息与计算机学院,太原030024

出　　处：《计算机科学》2021年第4期104-110,共7页Computer Science

基　　金：国家自然科学基金项目(61503273,61702356);教育部产学合作协同育人项目;山西省回国留学人员科研资助项目。

摘　　要：推荐系统是当前数据挖掘领域的研究热点,海量数据的涌现促使多源信息融合的推荐方法得到极大的关注。但是,现有的基于异质信息融合的推荐方法在进行特征表示时往往忽略了用户和项目之间的交互信息以及元路径之间的相互影响。因此,考虑到属性节点嵌入和结构元路径的不同视角,提出了一种多层次图注意力的网络推荐方法。该方法通过构建不同的元路径,将多源信息网络结构粒化为多个独立的粗粒度网络,然后基于图注意力机制结合局部节点属性嵌入,来分别学习用户和项目的潜在特征,最终给出融合后的细粒度网络推荐。在现实大规模数据集上进行横向和纵向评测,实验结果表明该方法能够有效地提升推荐性能。Recommendation system is a research hotspot in the field of data mining.Due to the emergence of massive data,the reco-mmendation methods of multi-source information fusion receive great attention.However,the existing recommendation methods based on heterogeneous information fusion often ignore the interaction information between users and items,as well as the interaction between meta-paths in feature representation.Therefore,considering the influence of different perspectives of attribute node embedding and structural meta-paths,a network recommendation method with multi-level graph attention is proposed.This method granulates the multi-source information network structure into multiple independent coarse-grained networks by constructing different meta-paths.Then,based on graph attention mechanism and local node attribute embedding,this method can learn the potential features of users and items separately.Finally,it gives a fine-grained network recommendation after fusion.The horizontal and vertical evaluations are conducted on real large-scale data sets,and the experimental results show that this method can effectively improve the recommendation performance.

关 键 词：层次粒化 多源信息融合 图注意力网络 Top-N推荐 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]