Parallel Community Detection For Large Scale Complex Networks Based On Conical Area Evolutionary Algorithm

In recent years,the community detection problem of complex networks has gradually at-tracted the attention of researchers in various fields.Community structure is one of the most important attributes of complex networks.The network community structure reveals important information of the network,which helps us understand and recognize the functional units of com-plex networks.The latest research on community detection methods based on multi-objective evolutionary algorithms shows that,optimizing multiple conflicting objective functions simul-taneously not only effectively overcomes some defects in single-objective optimization,such as the resolution limitation of modularity.On the other hand,community detection algorithms based on multi-objective evolutionary algorithms can also provide a set of hierarchical network community structures.However,the existing community detection algorithms based on the multi-objective evolutionary algorithms are only suitable for small networks.Therefore,it is of great significance to explore and study the detection problem of large-scale complex network communities.In this paper,the Conical Area Evolutionary Algorithm is first applied to complex network community detection problems.First,the community detection problem is modeled as a two-objective optimization problem and the excellent performance of the Conical Area Evolu-tionary Algorithm is fully utilized,and on this basis,more efficient parallel community detection algorithm is explored.The research work in this article includes(1)Model the community detection problem as a two-objective optimization problem with two conflicting objective functions,and propose Conical Area Community Detection.Its main framework is the Conical Area Evolutionary Algorithm(CAEA),which performs well on bi-objective optimization problems.Among them,CACD uses an efficient adjacency-based chromosome representation,which effectively reduces the solution space.In addition,the algorithm also uses the standard uniform crossover operator and uniform mutation operator suitable for this representation(2)This paper proposes Parallel Conical Area Community Detection Based on CACD,parallel global island model is designed,and elitist migration strategy is adopted to share important evolutionary achievements among the islands and to maintain the solution quality.In order to further improve the efficiency of the algorithm,PCACD also uses a process called net-work reduction.The original complex network will be converted into a small network with fewer nodes and edges.Therefore,the computational complexity of PCACD will decrease accordingly.(3)In the experimental part,this paper used several community detection algorithms based on multi-objective evolution algorithms as comparison algorithms,and the algorithm proposed in this article has four variants.Several artificial synthetic networks and real-world networks had been used to comprehensively evaluate the performance of each algorithm.Experimental results on artificially synthesized networks and real-world networks show that the algorithm designed in this paper not only outperforms other comparison algorithms in terms of solution quality and robustness,but also is more efficient for running time.