计算机应用研究2025，Vol.42Issue(11)：3275-3283,9.DOI:10.19734/j.issn.1001-3695.2024.11.0539
时空上下文感知的下一个PoI推荐方法
Approach for personalized next PoI recommendation for spatio-temporal context awareness
海燕 1王静 1刘志中2
作者信息
- 1. 华北水利水电大学信息工程学院,郑州 450045
- 2. 烟台大学计算机与控制工程学院,山东烟台 264005
- 折叠
摘要
Abstract
With the rapid development of location-based social networks,the next PoI recommendation has become a hot research topic in the recommendation field.However,existing research models ignore the spatio-temporal characteristics of PoIs and the effect of contextual information on the next PoI recommendation.To address this problems,this paper proposed an approach for personalized next PoI recommendation for spatio-temporal context awareness(STCNPR).Firstly,STCNPR used GAT to learn user representations containing social relationships.Then,STCNPR applied prevalence-enhanced bipartite graph neural networks(PEBGNN)to learn user representations containing PoI interaction preferences and PoI representations.Meanwhile,STCNPR applied ST-GCN to learn PoI representations containing PoI spatio-temporal transfer preferences.It finally fused the learned user and Pol representations to calculate the user's predicted score for each PoI and then recommend the next PoI to the user based on this score.To validate the method's effectiveness,this paper tested it on three publicly available datasets:Gowalla,Foursquare,and Yelp.The experimental results show that the proposed method demonstrates a significant advantage over several benchmark models in accuracy and recall,with an average improvement of 28.53％and 7.65％,respec-tively.关键词
下一个PoI推荐/PoI流行度/时空上下文/时空转移图/图注意力网络/时空图卷积网络Key words
next PoI recommendation/PoI prevalence/spatio-temporal context(STC)/spatio-temporal transfer graph/graph attention networks(GAT)/spatial temporal graph convolutional networks(ST-GCN)分类
计算机与自动化引用本文复制引用
海燕,王静,刘志中..时空上下文感知的下一个PoI推荐方法[J].计算机应用研究,2025,42(11):3275-3283,9.
