基于混沌粒子群的模糊C-均值聚类算法  

Research of fuzzy C-means clustering algorithm based on chaos particle swarm

作　　者：张春娜[1] 李轶然[1] 

机构地区：[1]辽宁科技大学软件学院,辽宁鞍山114051

出　　处：《计算机工程与设计》2013年第3期1039-1043,共5页Computer Engineering and Design

摘　　要：为了解决模糊C-均值(FCM)聚类算法的固有缺陷,提出基于混沌粒子群的模糊C-均值聚类算法(CPSO-FCM)。针对FCM对聚类初始值的敏感度问题,辅以粒子群算法以避免随机选取的聚类数和聚类中心所导致的结果不一致。通过引入混沌序列,在粒子的位置和速度上与原有粒子群优化算法所得计算值加以比较,取优者。这样不仅能够提高算法全局搜索能力,也可有助于粒子跳出局部最优。同时定义加速因子与逃逸算子对粒子移动速度加以优化,以加速收敛。实验结果表明,CSPO-FCM算法稳定性强,收敛速度快,且聚类的准确率高,效果较好。In order to solve inherent defect of the fuzzy Cmeans clustering algorithm （FCM）, this paper presents a fuzzy C means clustering algorithm （CPSOFCM） which based on chaos particle swarm. Firstly, according to the FCM on cluster initial value sensitivity problem, supplemented by particle swarm algorithm to avoid inconsistent results caused by the random selection of cluster number and cluster center. Moreover, by introducing a chaotic sequence, the particle position and velocity with the o riginal particle swarm optimization algorithm the calculation values can be compared, and the best. In this way, not only can im prove the algorithm global search ability, but also contribute to the escape of local optimal solution of particle. At the same time, because of definition of acceleration factor and escape operator on particle movement speed optimization, the convergence can be accelerated. Experiments prove that CSPOFCM algorithm has strong stability, fast convergence speed, high accuracy of cluste ring and better effect.

关 键 词：聚类 粒子群 混沌序列 模糊C-均值 CPSO-FCM算法 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]