Research On Node Influence Measurement And Influence Maximization Based On Topological Structure Of Networks

With the fast advancement of computer science and innovation and the expanding popularity of the utilize of the Web,online social network has gotten to be a crucial part of people's existence,and data dispersal is more closer on online social network.In the last few years,the research of social network has been broadly utilized in marketing,product promotion,viral marketing,public opinion control and other important fields.Therefore,it is of extraordinary viable noteworthiness to dig and analyze the gigantic data contained in social systems.The node influence measurement problem is through some indices to assess the importance of the nodes in the network,the influence maximization problem is by some way to dig out the k most influential seed node as a source of information transmission,makes the whole seed node as the original source of infection,under a given propagation model,the influence of transmission range is the largest,also is the foremost broadly the scope of data dispersal.However,the current strategies of estimating node influence have the disadvantages of high computational complexity and restricted application scope.How to proficiently and precisely solve the problem of influence maximization in large-scale networks is also a huge challenge.Therefore,this paper proposes a measurement method and some influence maximization algorithms based on this indicator combined with the method of network local topology.Our special works are as follows:First of all,this paper proposes the measurement method of node influence expectation.Through the analysis of the topology of the network around the node,that is,the propagation path of influence,the feasibility of limiting the influence to the thirdorder range is verified.This calculation method can quickly approximate the calculation of node influence,and use this method as an indicator to evaluate the importance of nodes,so as to find the nodes with high influence.Secondly,based on our calculation method of influence expectation proposed in previous chapter,we proposes an algorithm named LNS,which can quickly evaluate the influence of each node in the second-order neighborhood,and quickly get the k nodes with the greatest influence by ranking the influence of an individual node.However,what we need to solve is not to find the most influential node,but to solve the combination problem.We need to consider the overlapping of influence coverage.This paper proposes a common node neighborhood similarity strategy for the overlapping of influence propagation range of nodes to extend the LNS algorithm.In the final step,in order to measure the influence gain of nodes more accurately,the LNS?BS is improved by expected value reduction as LNS?BSDD algorithm,which makes the influence propagation range of the selected seed node set better than the existing heuristic algorithms.Thirdly,in order to solve the instability of the heuristic algorithms and meet the accuracy requirements of the strategy,this paper proposes an influence maximization algorithm LNS?CELF,which combines the community partition method and the influence expectation index to generate the candidate node set,so as to make greedy selection based on the candidate node set.The LNS?CELF algorithm can guarantee the accuracy and stability in theory as well as CELF algorithm,and further reduce the execution time of the algorithm,so that it can be used to quickly mining the most influential node sets in large-scale social systems.Finally,through the experimental verification in large-scale real network data set,the effectiveness of the strategy proposed by us is obtained,and the proposed multiple influence maximization algorithms based on the indicator has low computational complexity,but also can achieve a large spread range,so that they can be applied to large-scale networks.