Research On Influence Maximization Algorithm In Social Networks

The social media platforms such as We Chat and Tik Tok have shown a rapid development trend based on the development of the Internet.Everyone in the social network can share their thoughts and opinions with others instantly.It enriches people's spare time and promotes the exchange of information between humans.The social network is playing an unprecedented role.As one of the important issues in social network analysis,influence maximization is intended to find high-impact users.Identifying these high-impact users can provide decision-making assistance for enterprises and merchants in product marketing and other aspects,which is of great practical significance.The existing methods for measuring user influence generally rely on the topology of the network or the scope of information dissemination under the information dissemination models,which are mostly based on static networks.However,the dynamic nature of real social networks making it difficult to describe the true impact of users accurately for existing metrics.Considering the dynamic nature of real social networks,the existing theory of influence maximization is applied to dynamic networks to study the information dissemination mechanism and the calculation method of node influence in dynamic social networks.Based on the dynamic independent cascade model,a general formula for calculating node influence approximately is derived,and then a fast algorithm(Fast Iteration Algorithm)for calculating node influence is proposed.The experimental results under four real dynamic datasets show that the influence of the seed set selected by the influence maximization algorithm is higher than that of Forward Influence algorithm,Weighted Degree algorithm and Degree algorithm,and the time complexity of the influence maximization algorithm is lower.Besides,in some situations such as rumor control or disease transmission,we hope to prevent rumors or viruses from spreading by pretecting high-impact users in the network from misleading or contagious.Considering the change of network connectivity caused by node failures,this thesis maps the set selection problem of high-impact individuals in influence maximization onto the optimal percolation problem in networks.Then an influence maximization algorithm is proposed by taking robustness as an indicator.Besides,The efficiency of the proposed algorithm is optimized according to the monotonicity of the size of largest connected component.Discusses the effect of algorithm parameter on the robustness of the network and gives a reasonable range of values.The experimental results under eight real data sets show that compared with degree centrality,betweenness centrality and pagerank centrality,the influence maximization algorithm based on percolation can give the minimum percolation threshold of the network and the set of nodes to be removed.