Research On Influence Maximization Algorithm Based On Community Structure And Threshold

With the rapid development of the Internet,there are more and more applications with social attributes,such as the rightmost,interesting headlines and so on.The characteristics of these applications are huge number of users,interrelation among users,fast dissemination of information in these applications and wide range.It is very important to study viral marketing,public opinion analysis and traffic monitoring to maximize the influence.However,the existing heuristic algorithms for maximizing the influence do not take into account the overlapping influence of seed nodes.The problem of overlapping influence of seed nodes in the problem of maximizing influence is studied.Firstly,because the network community structure will affect the scope of information dissemination,a community structure-based maximum power algorithm CIMA(Community Influence Maximization Algorithm)is proposed.The algorithm defines community fitness and node contribution to delete sparsely connected nodes and select nodes closely related to the community,and divides the community on this basis.Based on the result of community division,the node with the largest degree of selection is used as seed node in each community.Secondly,for the problem of high time complexity of the problem of maximizing the influence,the influence maximization algorithm CTMD(Coverage Threshold Maximization Degree)based on the coverage threshold is proposed.The algorithm calculates the influence of all nodes in the network according to the improved influence estimation algorithm k-shell,selects the node with the most influence as the first seed node,and calculates the probability that the nodes within the two-level neighbor of the seed node set are activated.Based on the coverage threshold ?,the easily activated node is marked as the coverage state,and the node with the greatest influence and in the non-coverage state is selected as the seed node.Finally,the proposed algorithm is verified by experiments on artificial networks and real network datasets.The effects of CIMA,Degree(Maximal Degree)and IRIE(Influence Ranking Influence Estimation)are compared and analyzed.In the independent cascade IC and weight cascade WC models,the effects and running time of CTMA,CCA(Core Covering Algorithm),Degree and IRIE are compared and analyzed.