基于自适应Memetic算法的多目标复杂网络社区检测  

Multi-objective community detection in complex networks based on adaptive Memetic algorithm

作　　者：姚莹[1] 周井泉[1] 

机构地区：[1]南京邮电大学电子科学与工程学院,南京210003

出　　处：《计算机应用研究》2017年第3期858-861,共4页Application Research of Computers

基　　金：江苏省普通高校研究生科研创新计划项目(SJLX15_0377)

摘　　要：针对提高复杂网络社区检测准确度问题,提出了一种自适应Memetic算法的多目标社区检测算法。在全局搜索中利用Logistic函数来设置与全局优化相应的交叉概率和变异概率,并将多目标优化问题转换成同时最小优化kernel K-means和ratio cut这两个目标函数;在局部搜索中利用权重将两个目标函数合并成一个局部优化目标,并采用爬山搜索来寻找个体最优。在虚拟和真实网络实验平台下,与五种基于遗传算法的方法以及Fast Modularity算法相比,结果表明该算法能有效提高社区检测准确度,具有更好的寻优效果。In order to improve the accuracy of the community detection in complex networks,this paper proposed a multiobjective community detection based on adaptive memetic algorithm. In global search,the algorithm applied the Logistic function to set the corresponding crossover probability and mutation probability,and turned the multi-objective optimization problem into minimal optimization of two objectives called kernel K-means（ KKM） and ratio cut（ RC） at the same time. In local search,it constituted the local optimization target of weights of two objective functions and used a hill-climbing strategy to find the best individual. Experiments on synthetic and real life networks show that,compared with five algorithms based on GAs（ genetic algorithms） and Fast Modularity algorithm,the proposed algorithm can effectively improve the accuracy of the community detection and has certain advantages in solving community detection problems in complex networks.

关 键 词：复杂网络 社区检测 多目标 MEMETIC算法 自适应 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]