Social Networks on Big Data: Models and Dynamics

大數據的社交網絡: 模型和動力學

Student thesis: Doctoral Thesis

View graph of relations

Author(s)

Related Research Unit(s)

Detail(s)

Awarding Institution
Supervisors/Advisors
Award date4 Sep 2019

Abstract

The deluge of digital data on social networks offers unprecedented opportunities to investigate the mechanisms, dynamics and evolution of social networks. The study of social networks offers a deeper understanding of the connections among individuals across different geographic distances and diverse scales. And it provides significant insights into social ties, basic activity patterns and the underlying human behavior. Besides, the study of social networks, like Facebook, Flickr, Twitter and scientific collaboration networks, has witnessed significant progress recently, in which modeling the topological structures of the social networks and analyzing their dynamic properties have attracted most attention. To understand the processes through which social networks evolve, many network-generating mechanisms, such as self-organization of motifs, growth and preferential attachment, link prediction based on individual proximity, have been proposed. On the other hand, the tools and applications of information propagation on social networks more efficiently control the information spreading process, and address rumor problems. However, for information spreading over multiplex social networks, there are many unknown dynamics that are still attracting substantial attention from different research areas such as physics, mathematics and sociology. In this thesis, according to data observations, we explored the mechanisms and dynamic properties of social networks.

After our investigation, we found that tremendous empirical evidence in undirected social networks, such as Facebook, Quora and Foursquare, demonstrates that, to a large extent, individuals are associated with each other not by preference but through other organizing rules. One such rule found in many real social networks is the Henneberg growth mechanism. Based on this observed phenomenon, a Henneberg growth model is proposed in this thesis for modeling the Facebook network. Experimental results show that the function and structure of the model are in remarkable agreement with two huge-scale Facebook network datasets. The finding suggests that the Henneberg growth mechanism is fundamental to model some undirected social networks like Facebook.

The dynamic propagation of various information over social networks may gradually disseminate, drastically increase, strongly compete with each other, or slowly decrease. These observations had led to the present study of the spreading process of true and fake information over social networks, particularly Facebook. Through controlling the spreading parameters in extensive large-scale simulations, the final density of stiflers increases with the growth of the spreading rate, while it would decline with the increase of the removal rate. Moreover, the spreading process of the true-fake information is closely related to the node degrees on the network. Interestingly, it is found that the spreading rate of the true information but not of the fake information has a great effect on the information spreading process, reflecting the human nature in believing and spreading truths in social activities.

As for the multiplex social networks, we propose a competitive information model over social networks. The simulations of this model are verified by two types of multiplex networks, such as the real composite network and the artificial composite network. Through controlling the exchanging rate of competitive information, we are able to determine information dominance accurately. Then, we answer a question: how does the information spread on Facebook and Twitter? The analysis of the interrelation between two dynamic processes accounting for the information spreading, respectively preference and broadcasting. The analysis using a Markov chain approach reveals the phase diagram of the incidence of the two dynamic processes, and allows us to capture the spreading threshold depending on the topological structure of the multiplex network and the interrelation between the two processes. Interestingly, considering various information diffusion, we could determine information dominance by controlling the information incidence and the topology of the multiplex network.

In summary, we explore the mechanisms of the social networks, and investigate how to control the dynamic phenomena over social networks. We proposed mathematical models of social networks to reveal the human behaviors of the social platforms and fundamental mechanisms of social science.