Factors influencing message dissemination through social media

Zeyu Zheng, Huancheng Yang, Yang Fu, Dianzheng Fu, Boris Podobnik, and H. Eugene Stanley

Phys. Rev. E 97, 062306 – Published 7 June 2018

Abstract

Online social networks strongly impact our daily lives. An internet user (a “Netizen”) wants messages to be efficiently disseminated. The susceptible-infected-recovered (SIR) dissemination model is the traditional tool for exploring the spreading mechanism of information diffusion. We here test our SIR-based dissemination model on open and real-world data collected from Twitter. We locate and identify phase transitions in the message dissemination process. We find that message content is a stronger factor than the popularity of the sender. We also find that the probability that a message will be forwarded has a threshold that affects its ability to spread, and when the probability is above the threshold the message quickly achieves mass dissemination.

Received 28 November 2017
Revised 7 March 2018

DOI:https://doi.org/10.1103/PhysRevE.97.062306

Physics Subject Headings (PhySH)

Dynamics of networks Network phase transitions Network structure

NetworksNonlinear Dynamics

Authors & Affiliations

Zeyu Zheng^1,2,3, Huancheng Yang^3,1,2, Yang Fu^1,2,*, Dianzheng Fu^1,2, Boris Podobnik^4,5,6,7, and H. Eugene Stanley⁸

¹Shenyang Institute of Automation, Chinese Academy of Sciences, Shenyang, 110016, China
²Key Laboratory of Network Control System, Chinese Academy of Sciences, Shenyang, 110016, China
³University of Chinese Academy of Sciences, Beijing, 100049, China
⁴Luxembourg School of Business, Luxembourg, L-2453, EU
⁵Faculty of Information Studies, SI-8000 Novo Mesto, Slovenia, EU
⁶Faculty of Civil Engineering, University of Rijeka, 51000 Rijeka, Croatia, EU
⁷Zagreb School of Economics and Management, 10000 Zagreb, Croatia, EU
⁸Center for Polymer Studies, Boston University, Boston, Massachusetts 02215, USA

^*Corresponding author: fuyang@sia.cn

Article Text (Subscription Required)

Click to Expand

References (Subscription Required)

Click to Expand

Issue

Vol. 97, Iss. 6 — June 2018

Reuse & Permissions

Access Options

Article Available via CHORUS

Download Accepted Manuscript

Author publication services for translation and copyediting assistance advertisement

Images

Figure 1
Phase transition of page views and their distribution universally exist in the message dissemination. (a) Most of 1000 experiments rapidly rise within eight retweet steps. (b) Heavy line shows the fitting curve of page views which the page views rise quickly at the beginning, then start to level off. The light color interval is the standard error band of the fitting curves. (c) The green diamond shows the number of retweets while the blue triangle indicates page views, and they change synchronously. There are phase transitions exist in number of retweets and page views. More than 60% of messages can be spread widely. And others are hardly seen. (d) There are two distant peaks in the distribution of page views. Nearly 2/5 of the messages are hardly seen, and more than 3/5 of the messages could be spread across the network.
Reuse & Permissions
Figure 2
Impact of message's degree distribution on infection range is very little. (a) Degree distributions between hot users and normal users differ greatly. The average degree of hot users is more than 20 times beyond the average of normal users. (b) The probability density distributions of the infection ranges between the hot users and the normal users is not very different. Nearly two thirds of the messages sent by normal users get widely spread while all the messages sent by hot users spread well. So hot users cost a lot but gain less.
Reuse & Permissions
Figure 3
The effect on message propagation is getting bigger with retweet-probability $p$ rising. (a) It shows the page views curves of the spread process under three different retweet-probability. Each curve has a turning point that the curve's slope has decreased after the turning point. The bigger the $p$ is, the earlier the turning points appear. (b) It indicates that the total page views of messages increase while the retweet probability $p$ is increasing. The effect under different $p$ values on the total page views is not linear. In general, the bigger the $p$ value is, the smaller the influence on the infection range will be.
Reuse & Permissions
Figure 4
(a) The retweet steps needed for the same dissemination is approximately exponential in relation to the retweet probability $p$ . (b) Logarithmic curve of log( $p$ ) and log(steps) is divided into two sections. That means characteristics of network communication change with the two different slopes. The turning point of the curve is close to the one in the previous experiment. When the retweet probability increases beyond a threshold, the users who receive messages from different users have a great overlap, so that the page views increase more slowly.
Reuse & Permissions
Figure 5
The relationship between the maximum repetition number and the retweet-probability $p$ .
Reuse & Permissions
Figure 6
The relationship between log( $p$ ) and log(steps) under different scales of the network and different thresholds of “widely spread”.
Reuse & Permissions

Physical Review E

covering statistical, nonlinear, biological, and soft matter physics