文献类型：期刊文献

中文题名：基于粒子群优化的自组织特征映射神经网络及应用

英文题名：Self-organizing Feature Map Neural Network Based on Particle Swarm Optimizer and Its Application

作者：吕强[1];俞金寿[1]

机构：[1]华东理工大学自动化研究所,上海200237

年份：2005

卷号：20

期号：10

起止页码：1115

中文期刊名：控制与决策

外文期刊名：Control and Decision

收录：CSTPCD;;EI(收录号：2005489517194);Scopus;北大核心:【北大核心2004】；CSCD:【CSCD2011_2012】；

语种：中文

中文关键词：数据挖掘;自组织特征映射;粒子群算法;核函数;聚类

外文关键词：Data mining; Self-organizing map; Particle swarm optimizer; Kernel function; Clustering

摘要：采用粒子群优化(PSO)算法优化权重失真指数(LW D I),提出了基于粒子群优化的SOM(PSO-SOM)训练算法.用该算法取代K ohonen提出的启发式训练算法,同时引进核函数,以加强PSO-SOM算法的非线性聚类能力.以某工厂丙烯腈反应器数据为聚类应用研究对象,研究结果表明,与启发式训练算法相比,PSO-SOM算法能够得到较优的聚类,而且该算法实现简单、便于工程应用,对丙烯腈反应器参数调整以及收率监测具有显著的指导作用.
The self-organizing map （SOM） based on particle swarm optimizer （PSO）（called PSO-SOM） training algorithm is presented by using direct optimization of a locally weighted distortion index （LWDI） that is achieved through PSO algorithm. Kohonen＇s heuristic-based training algorithm is replaced by the PSO-SOM algorithm. Moreover, kernel mathod is introduced to strengthen performance of PSO-SOM nonlinear clustering. A real life application of PSO-SOM algorithm in classifying data of acrylonitrile reactor is provided. The experimental results show that this algorithm can obtain better clustering results than heuristic-based training algorithm and be easily applied for projects because of its simpleness.