Ms-hgat: information diffusion prediction based on memory enhanced sequence hypergraph attention network

This article introduces professor raoyuan's team from xi'anjiaotonguniversity 2022 Years published in AAAI An article on : “MS-HGAT: Memory-enhanced Sequential Hypergraph Attention Network for Information Diffusion Prediction”. Predicting the diffusion cascade is a key task to understand the spread of information on social networks . Here the author puts forward a new information diffusion prediction model —— Memory enhanced sequence hypergraph attention network (MS-HGAT). say concretely , In order to introduce the global dependencies of users , Not only did they make use of their friendship , And consider their interaction at the cascade level . Besides , To dynamically capture user preferences , The author divides the diffusion hypergraph into subgraphs based on timestamp , Development hypergraph Attention Networks To learn the sequence Hypergraph , And use the gated fusion strategy to connect them . Besides , A memory enhanced embedded lookup module is proposed , Capture the learned user representation into a cascade specific embedded space , Thus, the characteristic interaction inside the cascade is highlighted . The experimental results on four real datasets show that ,MS-HGAT stay [email protected] and [email protected] The indexes are obviously superior to the most advanced diffusion prediction model .

The article links ：

https://www.aaai.org/AAAI22Papers/AAAI-4362.SunL.pdf