Research On Heterogeneous Information Spreading Model In Complex Networks

Networks such as social networks, tranportation networks, power grid and so on have brought great convenience to people's life. Studies have shown that these large, complex net-works share surprising similarities including short average distance, highly-clustered and so on, which are known as complex networks. In recent years, the complex network has made a lot of progress, including the network modeling, performance analysis, network dynam-ics, synchronization and control, etc. In particular, information spreading gains widespread attention as it is of important theoretical significance and practical application value.Due to intensive study on information spreading in recent years, the propagation prob-ability and the nodes of the information spreading are heterogeneous, which is significantly different from the widely studied epidemic spreading. At the view of the propagation proba-bility, it is influenced by many factors, such as the tie strength, social reinforcement, memory effect and so on. At the view of nodes, they play various roles in information spreading due to their occupation, hobbies and so on. This dissertation focuses on how heterogeneity of links and nodes affects the spreading dynamics under the essential properties of information spreading. This dissertation consists of the following parts:Firstly, information spreading with heterogeneous transmission probability in complex networks is studied. Information spreading model in weighted networks is established and analyzed. Results shows that:certain clustering benefits information spreading, especially when tie strength tends to be homogeneous; and heterogeneity of tie strength seriously hin-ders information spreading. Based on this analysis, a tie strength calculation method is designed to adjust the transmission probability in complex networks, which improves the efficiency of the information spreading.Secondly, the identification of influential nodes in complex networks with heteroge-neous transmission probability is studied, i.e a influential nodes identification method based on Bayesian and Semiring Algebraic model. Different from the traditional network topology based methods for key node identification, this dissertation mines the interaction between the nodes and the transmission attenuation effect of node influence. Based on the interac-tion history between users, the link strength is calculated by a Bayesian estimation model with a forgetting discount. Then for individuals without direct interaction, the indirect link strength is obtained by the semiring model. Based on the small world properties of complex networks, the overall estimation of the user is achieved by the sum of both direct strength and the indirect strength. Finally, influential nodes are excavated.Thirdly, the ubiquitous node heterogeneity in complex networks is studied and a Graph Signal Processing based influential nodes identification method for complex networks with node heterogeneity is proposed. First, taking into account the role of nodes in information spreading, node attributes are classified and assigned corresponding coefficients. Then the graph signal is formed uniting the network topology. By Fourier transform, the main eigen-vectors are selected according to its spectrum and used to calculate the centrality of users. In this way, influential nodes are excavated comprehensively.This dissertation enlarges the understanding of the heterogeneous information spreading in complex networks. The results can be widely used in practical complex networks. It could be used to design network organization and spreading strategy to improves the efficiency of information spreading. Besides, node heterogeneity is in-vestigated in this dissertation, which will be helpful for further network analysis and control.