Abstract
Predicting the information spread tendency can help products recommendation and public opinion management. The existing information cascade prediction models are devoted to extract the chronological features from diffusion sequences but treat the diffusion sources as ordinary users. Diffusion source, the first user in the information cascade, can indicate the latent topic and diffusion pattern of an information item to mine user potential common interests, which facilitates information cascade prediction. In this paper, for modelling the abundant implicit semantics of diffusion sources in information cascade prediction, we propose a Diffusion Source latent Semantics-Fused cascade prediction framework, named DSSF. Specifically, we firstly apply diffusion sources embedding to model the special role of the source users. To learn the latent interaction between users and diffusion sources, we proposed a co-attention-based fusion gate which fuses the diffusion sources' latent semantics with user embedding. To address the challenge that the distribution of diffusion sources is long-tailed, we develop an adversarial training framework to transfer the semantics knowledge from head to tail sources. Finally, we conduct experiments on real-world datasets, and the results show that modelling the diffusion sources can significantly improve the prediction performance. Besides, this improvement is limited for the cascades from tail sources, and the adversarial framework can help.
Highlights
With the fast development of information technology and mobile communication technique, online media such as social networks and news platforms have become an indispensable way for people to acquire information and deeply influence the economy and culture
It is hard to learn the latent semantics of tail sources which is the majority in diffusion sources. To address these challenges above, in this paper, we propose Diffusion Source Semantics-Fused (DSSF), a diffusion sources latent semantics-fusion framework. e framework consists of three parts: firstly, for modelling the heterogeneity of different diffusion sources, we introduce diffusion sources embedding for each user who acts as a diffusion source
(2) Further, we find that the model with a co-attentionbased fusion gate performs better than the summation fusion model from the results shown in Figure 5. e results indicate that the interaction between users and diffusion sources is different from user to user, (3) from the results in Figure 6, we find that the adversarial training framework can effectively improve the performances of cascades from tail diffusion sources. e results show that the semantics knowledge in different diffusion sources is transferable, and we can solve the difficulty caused by data imbalance with adversarial learning framework
Summary
With the fast development of information technology and mobile communication technique, online media such as social networks and news platforms have become an indispensable way for people to acquire information and deeply influence the economy and culture. For public opinion government [2, 3] and marketing [4,5,6,7,8], it is important to mine the law behind information diffusion on online media [5]. Cascade-level prediction focuses on the global characteristics of the whole cascade such as the increment size [14, 17] and the popularity [9, 10, 18, 19]. User-level prediction focuses on the microuser behaviour in the network who will be the user retweet this post
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call