Research On Mining Algorithm Of Important Node Groups Based On Information Entropy

The iterative development of social networks has played a positive role in the rapid dissemination of information.Influence maximization research is widely used in commodity "viral" marketing,network monitoring,and rumor control.As a hot topic in the field of social network analysis,important node group mining has received more and more attention.Important node group mining refers to selecting a group of nodes from the network and spreading them under the propagation model to obtain the largest information dissemination range,which is a kind of social network influence maximization problem.However,in large-scale social networks,"rich-club" phenomenon leads to overlapping influences,making the maximization of the influence of node groups ineffective.This paper uses information entropy to measure the global and local importance of nodes,which can more accurately select nodes with strong information dissemination ability.This paper proposes a method to quantify the overlap of influence,adaptively updating the importance of nodes,which is beneficial to solve the rich-club problem.This research utilizes the community information of the network,can focus on the "bridge nodes",and further improve the efficiency of information dissemination.The main research work is as follows:(1)Combining the theoretical knowledge of information entropy,this paper designs a node importance evaluation index based on information entropy.Compared with classical indicators such as node’s degree,information entropy is conducive to the selection of initial node groups with greater influence and can evaluate the importance of nodes more accurately and scientifically.To make this indicator have good practical applicability,it is optimized for common static social network data,so that it can be applied to more application scenarios.(2)This paper proposes the overlap coefficient as a novel quantitative metric to quantify the overlap of influence caused by "rich-club" phenomenon.The overlap coefficient has a negative impact on the comprehensive influence of the node group.Combined with information entropy,this paper proposes an important node group mining algorithm based on information entropy and overlap coefficient.In this algorithm,information entropy is used to evaluate the global importance and local importance of nodes.For the influence of the neighbors around the most important node,the overlap coefficient is used to adaptively weaken,to avoid the phenomenon that the selected important nodes gather.(3)Combining the community properties of social networks,this paper proposes an important node group mining algorithm based on overlapping community discovery.To make the distribution of important nodes more widely,this algorithm first divides the network into multiple sub-networks,and then allocates the number of important nodes according to the contribution of each community to the entire network.At the same time,some overlapping "bridge nodes" of communities are selected as important nodes to improve the efficiency of information dissemination between communities.(4)Finally,the simulation experiment of the SIR propagation model is carried out using the above method.On the real network data,compared with the benchmarking algorithm,the average distance between the nodes selected by the algorithm in this paper is larger and the distribution is wider.The algorithms in this paper reduces the overlap of influence and obtains greater influence.