Research On Information Diffusion Based On Multi - Layer Social Network

Social network is used to describe the interrelationship among social individuals with specific social attribute. Nowadays, research on information diffusion in social networks has gained popularity in recent years. Taking the control of public opinion for example, information diffusion model is built to describe its propagation properties and underlying rules. Then, effective defense strategies are adopted to prevent its further propagation based the proposed model. Thus, how to build models describing information diffusion becomes a crucial problem. Along with the rapidly increasing scale of information, the social relationship among individuals becomes more and more complex, which incurs serious challenges to build effective diffusion models. Recently, Multi-layered Social Network has been proposed and great attention by researches to model the social topology and social attributes in a more realistic way, which abstracts the complex social relationships into a multi-layered social network with one type of social relationship in each layer.The basic idea to investigate information diffusion in social networks can be described as three steps. Firstly, the multi-layered social network is divided into different communities. Secondly, diffusion models are built based on these social communities. Finally, maximum diffusion goal function is proposed to describe the strategies of maximum influence. Although several models and algorithms have been proposed, however, almost all of these models cannot be used to investigate the information diffusion problem in MSN. Thus, this paper proposed a novel model and algorithm to deal with this problem.The main research work and contributions of this thesis can be generalized as follows:(1) Community detecting algorithm for MSN. Community is the cell of MSN. The relationship among one community is compact; however, the relationship among different communities is untight. The currently proposed community detecting algorithm mainly focus on one-layer social network and the detecting results is not accurate. Considering the multi-layered relationship, the diversity of different layers and connection strength among individuals, this thesis proposed a clustering-based multi-edge social network community discovery (CLEDCC), which considers the impact of social layers on information in social networks. The experiment results show that the proposed model and algorithms have high performance in accuracy and scalability compared with the previous works.(2) Information diffusion model for MSN. This thesis among the first presents the information diffusion model for MSN. The traditional models generally simplify the MSN into single-layered network and to conduct model analysis. However, since single-layered cannot effectively describe complex relationship among individuals, this thesis proposed a novel CRM model, which is based on the electric current streaming theory and incorporates Graph theory, probability theory and other mathematically theory. Theoretically analysis and experiment results show that the CRM is effective to model the underlying diffusion rules in multi-layered social networks.(3) Information diffusion. For an arbitrary social network (V, E), how to maximize the influence of information with least seed individuals is an NP-hard problem. The proposed CLEDCC algorithm can obtain high performance of community detecting results, more realistic diffusion process of information modeled by CRM, and high performance of maximum influence goal implemented by CDH-CLEDCC strategy.