| Benefit from the development of Internet technology,social networks have been deeply integrated into people’s daily lives.Social networks are profoundly changing and influencing people’s lifestyles,so it is particularly important to study them systematically in both academia and industry.Social networks are an important platform for information dissemination,and each individual in the network is exerted a certain degree of influence,and changes in the emotions,thoughts and behaviours of these individuals will affect the individuals around them.This model of maximizing benefits through word-of-mouth generated by target individuals is known as the viral marketing model,and the problem of finding these target individuals quickly and accurately in social networks is known as the influence maximization problem.Swarm Intelligence Optimization algorithms are currently a very popular research topic in the field of artificial intelligence.These algorithms are designed by simulating the swarming behaviour of birds,animals,insects and other organisms in nature,in order to achieve the purpose of optimization.Due to its excellent merit-seeking ability,it can be applied to the influence maximization problem to find the nodes with outstanding influence more accurately,but the swarm intelligence optimization algorithms solve continuous optimization problems and is not applicable to discrete social network spaces.This paper focuses on the combination of group intelligent optimization algorithms and the influence maximization problem,with the main contributions described below.(1)To address the problem that traditional influence maximization algorithms based on swarm intelligence optimization ignore the network structure in the initial node selection and the poor accuracy of the search node link,this paper proposes an influence maximization algorithm that draws on the motion of mayflies.Firstly,the motion model of mayflies in the mayfly algorithm is reconstructed in the node space,then an initial seeds selection strategy and a local search strategy based on the three-degree property are proposed,and both strategies are applied to the designed mayfly motion model.Finally,experiments were conducted on six real data sets and the algorithm showed excellent results in terms of propagation range and time efficiency.(2)To address the problem that traditional influence maximization algorithms based on swarm intelligence optimization have difficulty in handling the balance between global exploration and local exploitation,leading to the tendency to fall into local optimal solutions as well as consuming too much time,this paper proposes a cat colony-based influence maximization algorithm.This paper redesigned the strategy of the cat search and tracking model according to the characteristics of social networks.In the search mode,a node metric is proposed to determine the node contribution value in order to improve the identification of nodes.A search strategy based on propagation potential is used in the tracking mode to improve search efficiency in tracking mode.The state-of-the-art of the algorithm in this paper was validated on six real data sets.In this thesis,there are a total of 18 figures,6 tables,and 99 references. |
