Research On The Multi-objective Optimizing Methods For The Influence Spreading In Social Networks Based On Differential Evolution

With the continuous emergence of various social platforms and the explosive growth of massive data,online social networks play an important role in our daily communication and many other social activities,such as marketing activities of innovative products or services and cyberspace security governance,and its rapid development has attracted widespread attention.Influence Maximization is a hot topic in social network analysis,which aims to select κnodes from the network as the initial active set,and make them have the maximum range of influence propagation in the network by a specific information propagation model.In recent years,most of the work on influence maximization has focused only on how to maximize the spread of influence without focusing on other goals.However,in practice,influence maximization is not a simple single-objective optimization problem.Enterprises need to consider the cost factors when choosing individuals with great influence in marketing,and find influential individuals on the basis of minimizing budget costs to better meet the actual marketing needs.Therefore,the problem of multi-objective influence maximization is also worth our attention.To solve the above problems,based on the network topology structure,this paper proposes two algorithms to identify the set key nodes with wide influence spread coverage and low cost in the network,and the main research work is as follows:(1)A multi-objective discrete differential evolution algorithm with a fixed number of seed nodes κbased on multi-objective optimization is proposed to solve the problem of influence propagation maximization and cost constraint minimization in social networks simultaneously.Because each individual in a social network has different influence,the cost of being selected as a seed is also different.Firstly,based on the network topology,a budget cost constraint function is designed to measure the cost required by the seed set in the network,and the function and an influence propagation expectation function are formalized into a multi-objective influence maximization problem.Secondly,the degree centrality method is applied to the mutant population,where the node genes in chromosomes are sequenced and nodes with high degree are selected to replace those with low degree,allowing chromosome populations to find the Pareto non-dominant solutions with greater influence and propagation range more quickly.Finally,experiments are conducted on six real network datasets,and the results show that the method is effective in identifying influential nodes in the network and finding uniformly distributed Pareto non-dominated solutions.(2)A discrete multi-objective differential evolution algorithm with variable seed set size κis designed to solve the problem of maximizing the influence of multiple objectives.In practical marketing,enterprises may seek different number of influential users to promote products according to different needs,which leads to the need to consider the variability of seed set size when finding seed nodes in the network.The linear threshold model is used to simulate influence propagation,and the number of influence propagation and the number of seed nodes are taken as influence propagation function and cost constraint function respectively.In the population mutation operation,two mutation strategies are used to improve the global exploration and local exploitation performance of DMODE algorithm,and a strategy based on degree ranking is proposed to accelerate the convergence of the algorithm in this step.The algorithm is tested on four real network datasets,and the experimental results show that the Pareto non-dominant solutions found by DMODE algorithm has larger influence propagation and smaller cost,and the experiments verify that the two mutation strategies are more competitive than the one mutation strategy.