Abstract | ||
---|---|---|
Online social interactions have become more dynamic and far-reaching across multiple social media platforms than ever before. This is because of frequently changing online contact networks and increasing accessibility to diverse information sources outside of the networks. Accordingly, massive user-generated content spreads through heterogeneous social networks beyond a single social platform. The main goal of this paper is to propose a model-free approach for estimating macro-level information transfer across heterogeneous populations without any assumptions on such dynamic and complex social networks. With this approach, we estimate macro-level diffusion across mainstream news (News), social networking sites (SNS), and blogs (Blog), and the estimations are compared with outcomes from our previous model-driven approach. We also analyze crowd phenomena in diffusion for News, SNS, and Blog as online social systems in terms of activity, reactivity, and heterogeneity. We find that News is the most active, SNS is the most reactive, and Blog is the most persistent, which governs time-evolving heterogeneity. Discovered crowd phenomena are interpreted with respect to our proposed approaches. The strength and directionality of influence reflect reactivity, while topic-related diffusion patterns reflect heterogeneity. We expect that this study can provide a consistent way of understanding cross-population diffusion in diverse application domains. |
Year | DOI | Venue |
---|---|---|
2016 | 10.1016/j.neucom.2014.12.107 | Neurocomputing |
Keywords | DocType | Volume |
Macro-level information transfer,Transfer entropy,Crowd phenomena,Social media | Journal | 172 |
ISSN | Citations | PageRank |
0925-2312 | 0 | 0.34 |
References | Authors | |
5 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Minkyoung Kim | 1 | 85 | 11.01 |
David Newth | 2 | 77 | 22.00 |
Peter Christen | 3 | 1697 | 107.21 |