Font Size: a A A

Research And Application On Dynamic Spreading Model Based On Online Social Network

Posted on:2015-01-22Degree:MasterType:Thesis
Country:ChinaCandidate:Z Q MengFull Text:PDF
GTID:2268330428497429Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
The popularity of online social networks has not only transformed the way people communicate, but also enriched networks of people’s social relationships. Open and convenient features of the Internet bring convenience to people’s communication, meanwhile, they facilitate the dissemination of some gossip, rumors, viruses, etc. which are hard to control. Although current domestic and international researches on complex networks communication are developing rapidly, research on online social network as a new media faces enormous challenges. Studying laws of information spreading, analyzing mechanisms of public opinion propagation and identifying key factors of public opinion diffusion and influential nodes of network can effectively monitor the public opinion, and timely interrupt the diffusion of ill comments in online social networks. Towards to the information spreading issue, nowadays the academic circles mainly focus on modeling simulation which is constructed based on dynamics of complex networks, and some theories, models and methods from these dynamics are conducive to the better understanding of spreading behavior from different networks.However, traditional spreading models, because of its simple theory characteristic, are often difficult to describe real information dissemination process of online social networks. Therefore, the thesis firstly discusses the shortcomings of the mean field hypothesis and the contact-based annihilation mechanism in traditional models when simulating information spreading in online social networks. Then, we analyse features of online active behaviors, directed information spreading and inhomogeneous nodes in online social networks, and propose a dynamic spreading model(D-SIR) based on online social networks on the basis of a deep investigating and discussing the dynamics of complex networks. The D-SIR model introduces a annihilation function that spreaders can spontaneously switch their state to stiflers, and the core spreaders of network can avoid turning to stiflers in advance. The dynamic mechanism in this model that dynamically assigns authority and immunity to nodes allows model to descript the topological differences between nodes in online social networks. This model adopts directed graph as a communication network and can extend the influence factors of external social reinforcement when needed, all which can enhance applicability and scalability of the model.In the process of model validation, we analyze the basic topological features of three information spreading networks which are crawled from Sina microblog, and results show that there also exist the small-world and scale-free properties in these networks. Simulation results under different parameters which are simulated by using D-SIR model in these real information spreading networks are conformity with the reality of propagation process in online social networks. Finally, in order to apply our model to identify influential nodes in networks, we exam the spreading influence of every node by the simulative spreading originating from a single node in real networks. The analysis of the correlation between the spreading influence and the centralities of nodes reveals that spreading capabilities of the nodes do not depend on their k-core index, instead, the out-degree and the closeness centrality seem to be the better topological descriptor to locate such influential individuals. Our finding is helpful for identifying key nodes in the information dissemination networks, and also lays a foundation for further studying the spreading mechanisms of information and rumors in online social networks.
Keywords/Search Tags:Modeling simulation, Spreading model, Dynamics of Complex networks, Online social networks, Influential nodes
PDF Full Text Request
Related items