网络社区检测的一种新型混合算法  

A network of community testing new hybrid algorithm

作　　者：路延[1] 

机构地区：[1]陕西职业技术学院,710100

出　　处：《电子测试》2014年第8期32-34,共3页Electronic Test

摘　　要：本文中介绍结合粒子群优化算法和非负矩阵分解算法,得到了一种新型社区检测的谱聚类算法PNMF-Net。这个算法将模块度作为目标函数。由于PSO算法容易陷入局部最优,对粒子群算法中个体历史最优粒子使用NMF搜索。同时NMF算法可以减小搜索空间,自动找到网络社区数。最后使用人工合成网络和四组真实网络对PNMF-Net进行测试,结果证明提出的谱聚类算法可以有效地检测出网络中固有的社区。In this paper,a novel spectral algorithm（PNMF-Net）to detect communities by combining the algorithms of Particle Swarm Optimization（PSO）and Nonnegative Matrix Factorization（NMF）is proposed.The proposed method optimizes a quality function known as modularity.As PSO may prematurely converge on local optimal solutions,NMF is applied for exploitation by locally modified some better particles resulted by PSO.Meanwhile,NMF narrow the searching space and automatically detect the number of communities of networks.Finally experiments on artificial and real world networks are presented to show the effectiveness of the proposed method.

关 键 词：社区检测 谱聚类 粒子群优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]