Research On Ranking And Measuring The Influence Of Vital Nodes In Complex Networks

Over the last decade,the prominent researches and the enormous developments of the complex networks theory provides people with a novel perspective and new tool on the reality world.Actually,almost all the complex systems of the real world,such as social,information,biology,economic and financial,and electrical and transportation,which according to abstract processing in both time and space dimensions,can be modeled and expressed as network structures naturally.Furthermore,the theories of the network science are usually employed to clearly describe and solve problems of complex systems.The vital nodes in complex networks are the extraordinary nodes which play more significant and central roles than other nodes on the structure and function of the complex networks.Obviously,the mining of the vital nodes in the complex networks are closely related with the spreading,synchronization,controllability of the networks,and their evaluation methods have been proved to have very important significant to improve robustness and reliability of networks.Therefore,for the great theoretical significance and the practical value have been show in the researches of this field,which have been a hot topic in the field of complex networks.This thesis introduces the two important topics of the mining of the vital nodes in complex networks: identifying and ranking important nodes;maximizing the influence of nodes.We mainly carried out the following work:Firstly,we inspired by the local properties of the network and proposed a novel centrality named Clustered Local Degree(CLD),which combines the calculated sum and the clustering coefficient of nodes to evaluate and sort the propagation capabilities of all nodes in the network.Extensive simulations are performed on a series of real networks,the results show that the CLD centrality exposes a competitive performance in distinguishing the spreading ability of nodes,and exposes the best performance in the accuracy of ranking spreaders.Secondly,we conduct the research on community-based influence maximization by proposing a new heuristic framework to find the initial seed nodes set.After the network is divided into communities,the initial seed nodes set is selected and propagated in each community using the CLD method,and propagated with the set of initial seed nodes sorted by the CLD method.A large number of simulations were conducted on the classic social network propagation models IC and LT,and the results shows that the seed node set selected by the CLD method after being divided by the community can affect more nodes in the network,and the information is more likely to spread when the nodes' active threshold is identical.It is worth mentioning that the CLD is a parameterfree measure with a competitive computational complexity,which is suitable to apply on the relatively dense networks.