Research On Influence Maximization In Dynamic Complex Networks

In recent years,with the rapid development of computer technology,social network has been widely used.In social networks,people can make new friends and share information,which provides great convenience for the Internet users.In the process of popularization and application of social network,the problem of influence maximization arises.It is a hot and difficult problem in complex network research because of its wide application and great development prospect.In order to solve the problem of influence maximization,many experts and scholars have proposed greedy algorithm and heuristic algorithms,which can solve the problem well to some extent.However,with the gradual expansion of network scale,the constant change of network topology,and the frequent increase and decrease of users in the network,the research focus of influence maximization has gradually shifted from static network to dynamic network.The problem of influence maximization in dynamic complex networks needs to be considered from three aspects: first,how to select the appropriate influence propagation model,and there will be different choices for different network structures.Second,how to accurately assess node influence,it is mainly based on some centrality indicators.Third,how to select the seed nodes should not only ensure the final influence is large enough,but also not waste too much time.In view of the above problems,this paper combines the existing influence maximization algorithms and conducts analysis and research from the following three aspects:(1)This paper refers to the concept of entropy.The concept of propagation entropy is proposed to evaluate the influence of nodes.Compared with degree centrality,betweenness centrality and closeness centrality,propagation entropy not only considers the influence of the node itself,but also pays more attention to the influence of node location and multi-order neighbors.And it is verified in a small network that the propagation entropy can more accurately evaluate the influence and importance of a node,and the influence of a node can be ordered more accurately.(2)A heuristic algorithm based on propagation entropy is proposed.The algorithm selects the node with the maximum propagation entropy in the central core layer of the current network as the seed node,the degree of common neighbors of candidate nodes and seed nodes is calculated,some nodes are discounted and some nodes are overwritten,the nodes with less common neighbors and higher propagation entropy are selected as new seed nodes.This algorithm not only selects nodes with high influence as seed nodes,but also ensures that the seed nodes are sufficiently dispersed,reduces the overlap of influence,and makes the influence propagation range large enough.(3)We selected infectious disease model,experimented on four networks,and compared with four algorithms of influence maximization problem.The algorithm proposed in this paper can select the set of seed nodes with a large influence propagation range in a relatively short time.It is proved that the algorithm is effective,feasible and advanced.