Comprehensive Evaluation Method For Propagating Sources In Complex Networks

Under the complicated social network,a large amount of false information exists in the network,which harms the interests of people.At the same time,rapid urbanization and convenient transportation make the communication between people more frequent,which brings rapid spread of infectious diseases.Therefore,it is of great significance to find and control the source of diffusion events.The propagation source point in a complex network forms an infection network after a certain period of time.In this thesis,the limited information of the infection network at a certain time,such as the infection status of the node and the incomplete propagation path,is used to find the true source point.The difficulty of the source point problem is that the complete propagation path or the infection status of all nodes may not be obtained through a complex network.Therefore,using the basic network and the infection network,this thesis proposes a source point mining method that combines the propagation subgraph and observation points.As for the deployment strategy of observation points,this thesis uses an improved k-shell decomposition method combined with Jordan’s central idea to perform observation point distribution.At the same time,based on this,combined with the evaluation value of the propagation subgraph,we jointly estimate the possibility of infected nodes as the source point.Then based on this,the single-source point model is extended to study the multi-source point mining combined with the single-source algorithm.This thesis simulates the information transmission of single or multiple source points by constructing a network and a real complex network,and finally acquires the infection subgraph after a certain time.Experiments were performed on a variety of different network diagrams,and the feasibility and accuracy of the single-source algorithm for comprehensive evaluation proposed in this paper was studied,and compared with various single-source algorithms.Then,different observation point deployment strategies were adopted.The superiority of the observation point strategy proposed in this paper is verified.Finally,the single-source point model is extended to an adaptive multi-source point algorithm.The feasibility and accuracy of the multisource algorithm are verified through experiments and comparison with the CMS model(based on community detection)under real and constructed networks.