We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.LG

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Computer Science > Machine Learning

Title: Independent Asymmetric Embedding for Information Diffusion Prediction on Social Networks

Abstract: The prediction for information diffusion on social networks has great practical significance in marketing and public opinion control. It aims to predict the individuals who will potentially repost the message on the social network. One type of method is based on demographics, complex networks and other prior knowledge to establish an interpretable model to simulate and predict the propagation process, while the other type of method is completely data-driven and maps the nodes to a latent space for propagation prediction. Existing latent space design and embedding methods lack consideration for the intervene among users. In this paper, we propose an independent asymmetric embedding method to embed each individual into one latent influence space and multiple latent susceptibility spaces. Based on the similarity between information diffusion and heat diffusion phenomenon, the heat diffusion kernel is exploited in our model and establishes the embedding rules. Furthermore, our method captures the co-occurrence regulation of user combinations in cascades to improve the calculating effectiveness. The results of extensive experiments conducted on real-world datasets verify both the predictive accuracy and cost-effectiveness of our approach.
Subjects: Machine Learning (cs.LG); Social and Information Networks (cs.SI)
Journal reference: 2022 IEEE 25th International Conference on Computer Supported Cooperative Work in Design (CSCWD)
DOI: 10.1109/CSCWD54268.2022.9776071
Cite as: arXiv:2105.08291 [cs.LG]
  (or arXiv:2105.08291v6 [cs.LG] for this version)

Submission history

From: Wenjin Xie [view email]
[v1] Tue, 18 May 2021 05:40:38 GMT (1039kb,D)
[v2] Thu, 17 Jun 2021 01:32:50 GMT (1050kb,D)
[v3] Wed, 24 Nov 2021 09:21:54 GMT (1789kb,D)
[v4] Wed, 1 Dec 2021 12:29:35 GMT (1789kb,D)
[v5] Tue, 1 Feb 2022 15:42:46 GMT (1790kb,D)
[v6] Thu, 26 May 2022 08:29:48 GMT (1788kb,D)

Link back to: arXiv, form interface, contact.