Research Of Community Detection Algorithm Based On Discrete Grey Wolf Optimization

A cluster of closely connected nodes which have same attributes or functions in a complex network is called a community.The complex network can be divided into several communities,and the community structure is an important feature of complex networks.The topological structure and module function of a network can be further addressed to solve real-life issues by analyzing the community structure.Therefore,many community detection algorithms have been proposed recently,such as spectral method,label propagation based methods,random walk based methods,network dynamics based methods,optimization,and so forth based methods.Among them,the community detection algorithms based on heuristic algorithm can achieve good performances.By analyzing the shortcomings of this kind of algorithms and the traditional algorithms,this paper proposes two community detection algorithms based on grey wolf optimization(GWO):(1)GWO-net is an algorithm based on both of discrete grey Wolf optimization framework,and the idea of label propagation.The framework uses a kind of discrete linear coding,which can indicate the non overlapping community structure.In order for the Grey wolf optimization to apply to community detection,GWO-net redefines the formula and operation of GWO.The algorithm uses the function named Modularity Density with a parameter.Under different parameters,GWO-net gets the different levels of community structure.In the initial phase,the algorithm uses two initializing methods based on the label propagation.One method effects on the global network,and the other effects on the local of nodes.The mountain search strategy is used to explore the surrounding areas of the best search agent in each iteration.In the later period of evolution,some search nodes will be randomly selected to mutate.To verify the performance of the algorithm,we compare GWO-net with other non overlapping community detection algorithms.It can be seen that the GWO-net has significant advantages compared with other algorithms.(2)OLGWO-net is proposed to solve the problem of overlapping community detection,which is based on discrete grey wolf optimization framework,and the contact information of the neighborhood associations of nodes.Since the algorithm uses a nolinear coding,make the search agent can store multiple community labels.The algorithm uses the extension modularity as the fitness function to evaluate the overlapping community structure.To determine a overlapping node,OLGWO-net uses the sum of node's neighborhood communities inner degree and the standard deviation.The sum can evaluate the tightness of the relation between the node and the neighborhood communities.In the next period of evolution,OLGWO-net uses the mutate operation to keep the diversity of the search agent swarm.The algorithm uses the mountain search strategy which has two situations that tend to get the non overlapping community structure and the overlapping community structure.To evaluate the performance of the algorithm,we compare the OLGWO-net with other some overlapping community detection algorithms.It can be seen that the OLGWO-net has strong community detection ability.