中文摘要：

文中研究复杂网络社区检测机制,提出了一种基于Memetic算法的多目标社区检测算法。为了提高种群多样性、减少搜索空间和提高算法效率,算法采用标签启发式快速传播的初始化策略,混合交叉,在每个社区中选择一个节点变异等优化两个目标函数,即Improved Ratio Association（IRA）和Ratio Cut（RC）,将多目标优化问题转化成同时最小优化这两个目标函数;在局部搜索中利用权重和将两个目标函数构成一个局部优化目标并采用爬山搜索来寻找个体最优。针对计算机合成网络与两个经典真实网络的实验结果表明,与四个基于EA的算法和Fast modularity算法相比,基于Memetic算法的多目标复杂网络社区检测机制在解决复杂网络社区检测问题上具有一定优势。

英文摘要：

The complex network community detection mechanism was studied and a multi- objective community detection based on M emetic algorithm was presented. In order to improve the diversity of the population,reduce the search space and raise the efficiency of the algorithm,the initialization strategy of label heuristic fast propagation and hybrid crossover were used in the algorithm and a node was selected in each community for mutation to optimize two objective functions,namely Improved Ratio Association（ IRA） and Ratio Cut（ RC）,which turns the multi- objective optimization problem into minimal optimization of these two objectives at the same time. In local search,the local optimization target is constituted of weights of two objective functions and a hill- climbing strategy is used to find the best individual. Experiments on computer- generated networks and two classic real networks showthat compared with four algorithms based on EAs and fast modularity algorithm,multi- objective community detection based on M emetic algorithm has certain advantages in solving complex network community detection problem.