In this work, we study the event occurrences of user activities on online social network platforms. To characterize the social activity interactions among network users, we propose a network group Hawkes (NGH) process model. Particularly, the observe
d network structure information is employed to model the users dynamic posting behaviors. Furthermore, the users are clustered into latent groups according to their dynamic behavior patterns. To estimate the model, a constraint maximum likelihood approach is proposed. Theoretically, we establish the consistency and asymptotic normality of the estimators. In addition, we show that the group memberships can be identified consistently. To conduct estimation, a branching representation structure is firstly introduced, and a stochastic EM (StEM) algorithm is developed to tackle the computational problem. Lastly, we apply the proposed method to a social network data collected from Sina Weibo, and identify the infuential network users as an interesting application.
In this article, we study the activity patterns of modern social media users on platforms such as Twitter and Facebook. To characterize the complex patterns we observe in users interactions with social media, we describe a new class of point process
models. The components in the model have straightforward interpretations and can thus provide meaningful insights into user activity patterns. A composite likelihood approach and a composite EM estimation procedure are developed to overcome the challenges that arise in parameter estimation. Using the proposed method, we analyze Donald Trumps Twitter data and study if and how his tweeting behavior evolved before, during and after the presidential campaign. Additionally, we analyze a large-scale social media data from Sina Weibo and identify interesting groups of users with distinct behaviors; in this analysis, we also discuss the effect of social ties on a users online content generating behavior.
