.png){kind=icon}
返回
Multi-session aware hypergraph neural network for session-based recommendation
.png){kind=icon}
PRE
AI
DOI：10.1007/s11042-023-15894-w.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Session-based user behavior prediction is a difficulty in network behavior modeling due to the limitation of information. In recent years, the neural network has become a new research direction in recommendation system, however, the existing graph structure recommended method simple binary relation of concern within the session, but in real life tend to have the multiple complex relationships between items. In addition, hyperedges lack displayed position information in hypergraphs, and items in different orders may construct the same hyperedges, which necessarily limits the ability to obtain exact vector representations of sessions. Therefore, to solve the above limitations, a multi-session aware hypergraph neural network (MA-HGNN) for session-based recommendation is proposed, which takes advantage of hypergraphs to model complex multivariate relationships in sessions, and alleviates the hyperedge isomorphism problem by preserving sequence information. At the same time, the co-occurrence graph structure and the local session graph structure are established to realize the connection between the similar user intentions in different sessions and the potential behavior patterns in the same session. Finally, experiments are carried out on three real-world datasets Diginetica, Tmall and Nowplaying, and the models proposed in our work are significantly improved, which proves the effectiveness of the method.
Keyword:
Multi-session aware
Soft attention
Hypergraph neural network
Session-based recommendation
期刊
IF:
3
论文数:
2.0W
被引数:
3.2W
机构
引用论文
分享收藏Conjugated heat transfer and temperature distributions in a gas turbine combustion liner under base-load operation基本负荷运行下燃气轮机燃烧衬里中的共轭传热和温度分布
PREAI
分享收藏
Individual and combining effects of anti-RANKL monoclonal antibody and teriparatide in ovariectomized mice.png)
Bone Reports
OAAI
Bone ReportsIF0
OAAI分享收藏
查看更多内容