Research On The Influential Nodes Selecting Mechanisms Based On Particle Swarm Optimization In Social Networks

The rapid development of information and Internet technology has changed people’s working,studying and living styles,and it also has a notable impact on the economic activities.Social networks,consisting of a large number of social individuals and organizations and their complex interactions with each other,have become important platforms for activities such as tracing the source of information dissemination,guiding the direction of public opinion,developing the network economy,carrying out social governance,and maintaining national security,etc.An individual in the network often has different influences due to the role of perceptions,culture,emotions,hobbies and other factors.In the process of information dissemination,the nodes with high influence will reshape the cognition and behavior of their neighboring nodes,and this influence cascade propagation leads to the continuous dynamic evolution of social network topology.As one of the important research contents of social network analysis and social computing,influence maximization problem aims to select a certain number of nodes as the seed set in a given network,and make the seed set spread the influence in the network to the widest extent under a specific propagation model.The study of influence maximization can not only enrich the research on the theoretical aspects of complex networks,but also has some application value in practical scenarios such as viral marketing and rumor control.Existing algorithms for solving influence maximization problems in networks of different sizes often sacrifice solution accuracy or computational cost,and are not well suited for networks of different structural types.To this end,the feasibility and advantages of the metaheuristic algorithm in solving the influence maximization problem are thoroughly investigated in this paper;the effects of the global search strategy and local exploration strategy of the discrete particle swarm algorithm on the solution performance of the algorithm are studied for the sake of the algorithm’s solving ability in the network;on this basis,the fast evolution mechanism of the discrete quantum particle swarm algorithm and Levy flight is studied based on the network topology.Experiments show that the seed selection strategy based on the particle swarm optimization algorithm is very effective in solving the influence maximization problem.The specific work of this paper is summarized as follows:(1)To improve the algorithm’s ability in identifying influential nodes in the network,an improved discrete particle swarm algorithm is proposed based on a discrete particle swarm algorithm that investigates the role of the strategies of global and local search in improving the solution quality of the algorithm.To this purpose,the feasibility as well as the advantages of the metaheuristic algorithm in solving the influence maximization problem are thoroughly investigated in this paper.When the classical particle swarm algorithm is used to deal with the influence maximization problem,a strategy to identify the particles falling into local optimum is constructed,and the idea of unbiased search is introduced in the local search,so that the particles falling into local optimum prematurely as well as the globally optimal particles can be searched in the network space better.Experiments show that the performance of IDPSO has been improved to some extent.(2)Based on the above work,the evolutionary mechanisms of the discrete quantum behavior-based particle swarm algorithm and Levy flight are combined with the characteristics of the network structure,and an attempt is made to apply these two algorithms to solve the influence maximization problem.Since the particle swarm optimization algorithm based on quantum behavior converges quickly and easily falls into local optimum in solving the problem,the proposed Levy flight algorithm based on the shortest path candidate nodes pool selection strategy enables the algorithm to converge to a certain range in a short time before expanding the search range by Levy flight,so as to efficiently identify the set of influential nodes in the network.Experiments show that the algorithm can achieve a solution performance very similar to that of the CELF algorithm based on the greedy strategy.