Research On Community Detection And Propagation Dynamics In Complex Networks

Complex systems in the real world can be described by complex networks which have some or all of the features:small-world,scale-free,assortativity or disassortativity,hierarchical structure,community structure.The classical propagation model only focuses on the process of change between different states or types of nodes,regardless of the influence of network structure.The study of propagation dynamics based on complex networks pays more attention to the influence of network structure,personal behavior,propagation characteristics and evolution.The research of network structure extends from node to cluster,community and hierarchical structure,and describes the internal relationship of complex system more accurately from different angles.Community structure is characterized by the tight connection of nodes in the module,but the connection between modules is sparse.Humans are carriers of viruses.The propagation chain formed by people's behavior usually has the characteristics of community structure.However,it is not completely clear what factors affect each other between community structure and propagation dynamics.Most of the existing complex network models are based on regular models,which is not consistent with the networks formed in the course of human epidemic transmission.In this paper,the structure and the robustness of the community,and according to the characteristics of the epidemic spreading and the behavior of people,the establishment of propagation model close to reality and the design of effective prevention and control strategies are studied.The specific work is introduced as follows:(1)A community structure detection algorithm based on resource allocation of bipartite networks,termed ECD.According to sexually transmitted diseases with dichotomy and community structure,how to accurately identify modules and their boundaries in bipartite networks is studied to improve the quality of community detection.The nodes are divided into two categories where the same kind of nodes has no intersection and the different nodes are related to each other.Through the projection based on extending resource allocation,we find not only the weights between nodes of the same type,but also the weights between nodes and their neighboring communities.The redundant edges which are connected to low-weight neighboring communities are found.ECD algorithm is used to detect overlapping nodes and community structure.Compared with the experimental results of artificial and real networks,ECD algorithm has excellent performance.In addition,the edge is the main medium in propagation,and overlapping nodes may promote the spreading of viruses among the communities.Removing redundant edges and taking active prevention and control measures on the overlapping nodes can inhibit the spreading of this virus to other communities.(2)Assortative neighbors classification priority link coupling algorithm is proposed,termed ANPL.The real networks are not isolated,but there is a coupling of various modes.How to improve the community robustness in interdependent networks by keeping the subnet structure unchanged is studied.We focus on the factors that the characteristics of community structure,degree correlation of nodes and mechanism of cascading failures in interdependent networks.The process of cascading failure is simulated on the generated two-layer interdependent networks coupled by scale-free networks.The robustness and similarity of community are analyzed by changing the ratio and location of failure nodes,and community structure strength.Compared with the three classical coupling algorithms,the experimental results verify the effectiveness of ANPL algorithm.This conclusion provides guidance for further study of multi-layer interdependent networks model with strong community robustness.(3)A novel epidemic spreading model with dual quarantine strategies in scale-free networks with overlapping community structures is proposed,termed SIQS.In view of influenza and coronavirus transmission in populations,it not only considers that the susceptible individuals face different infection risks in their own communities,but also analyzes the virus characteristics.The state of infected persons is divided into incubation period and outbreak period.An individual in incubation period is also infectious,and the isolation rate should change dynamically with different stages of individual infection.By changing the parameter values to adjust the isolation strategy,SIQS model is simulated on scale-free networks with overlapping community structures.The experimental results show that the isolation density of dual quarantine strategies is smaller than that of the strategy for isolating the infected individuals.And it can reduce speed of epidemic spreading from local to global.The effect is related to the latency,infection rate and initial infection density.The results of experiments verify the validity of the model.And it provides a theoretical basis for the further design of reasonable prevention and control strategies.(4)The homogeneity of epidemic spreading in complex networks is studied.In the study of epidemic spreading dynamics in complex networks,the analytical solution is usually obtained by using the hypothesis of uniform mixing,that is,the disease is uniformly distributed in the networks.However,only local contact between individuals leads to the spreading of disease in social networks.By analyzing the deficiency of the characteristic infected cluster size,the normalized entropy of infected cluster size(?*)is proposed to measure the change of all infected clusters.In the simulation experiments,the SIS model is used in the dynamic networks with long-range and local random walks.By comparing ?*in the dynamic networks with the homogeneous models,the results show that homogeneity of epidemic distribution is influenced by the moving speed of individual,infection density,infection rate,long-range motion probability and action radius.To sum up,the focuses of researches on different network structures are different.In order to establish a more realistic transmission model and design a reasonable and effective prevention and control strategy,the interaction between network structure,virus transmission characteristics and behavior pattern should be fully considered.