一种改进的多目标社区结构检测算法

An Improved Algorithm for Multi-Object Community Structure Detection

社区结构是复杂网络的重要特征之一,社区结构的检测也日益受到研究者们的关注。针对基于非支配多目标社区检测算法(MOGA-NET)在多样性方面存在不足的问题,提出一种改进的多目标社区结构检测算法(ICDMOGA-NET)。该算法将社区检测问题建模成多目标优化问题,结合向量编码方式、双向交叉算子与统一变异算子,对MOGA-NET算法进行改进。通过与原始的MOGA-NET算法进行比较,该算法在空手道俱乐部真实网络上的模块度Q以及归一互信息NMI分别大约提高3.01%、12.63%,对海豚社交网络的平均模块度Q大约提高11.3%,因此所提算法可以提高小型网络社区检测的准确率及稳定性。

Community structure is one of the important features of complex networks,and its detection is also attracting increasing attention.However,there is a shortcoming of diversity for non-dominant multi-objective genetic algorithm(MOGA-NET).In order to solve the problem,an improved community detection algorithm(ICDMOGA-NET)is proposed.The proposed algorithm based on MOGA-NET,models the community detection problem as a multi-objective optimization problem.In the process of evolution,it combines the vector encoding method,the two-way crossover method and the unified mutation operator.By comparing with MOGA-NET algorithm,the modularity Q and normalized mutual information(NMI)of the proposed algorithm on the real network of the karate club increase by 3.01%and 12.63%,respectively.The average modularity Q on the dolphin social network increases by 11.3%.Therefore,the proposed algorithm has higher accuracy and better stability for community detection in small network.