Dynamic Community Detection And Evolution Analysis Based On Topology Potential

Dynamic community detection and evolution analysis are two widely concerned areas of complex network.Current methods of dynamic community detection generally update the modular structure by recalculating the exact incremental parts of the network.Compared to recalculating the structure from the scratch and reassigning the affiliation of all nodes,the strategies based on incremental reprocessing can significantly improve the efficiency.However,nodes in complex networks have an influence on the neighbors around them,and limit the scope exactly contain the changed nodes when update the community structure may lead to the deviation of the results of community detection while these errors will accumulate from timestep to timestep.As to community evolution analysis,most of the current methods are over-reliance on external parameters,thus reduce applicability of the algorithm,and there are some other methods of community evolution are of complex mechanism,and cause the difficulties to apply them.This paper presents a dynamic community detection method based on incremental local updating.Firstly,it detects the initial community structure according to the positions of nodes in the inherent peak-valley structure of the topology potential field to improve the accuracy of the assignments of nodes.In addition,we take the great consideration of the interaction among nodes,and extend the local-update around the affected boundary of the incremental nodes.Experiments show that our method is able to deliver a good division of community structure and can effectively avoid error accumulation which traditional local-update algorithms will be when the network suffers significantly changes.This paper presents a community evolution analysis method based on core nodes.We take the advantage of the characteristic that the core nodes are relatively stable in community,and track the community evolution by detecting the change of the core nodes at every time step and thus simplifies the mechanism when analyzing the evolutionary of communities.Based on the theory of topology,this method search the peak node from the topology potential field as the core nodes of each community,and analyze the number as well as the transformation of attribute of core nodes according to the characteristics of the peak nodes at each time step.Experiments show that this method performs well in tracking the evolution events of communities.Moreover,it's free parameters,which improved the applicability of the method.