arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 1402.6428cs.NE
收藏

基于PSO的多维数据聚类算法

Clustering Multidimensional Data with PSO based Algorithm

Jayshree Ghorpade-Aher, Vishakha A. Metre

摘要:

数据聚类是数据采掘中公认的数据分析方法,而K-Means是众所周知的分区聚类方法,具有令人愉快的特性。我们观察到,K-Means和其他分区聚类技术受到一些限制,如初始聚类中心的选择,预先知道的集群数量,死单元问题,多个集群成员和过早收敛到局部最优。为了解决聚类的局限性,已有多种优化方法被提出,但群体智能(SI)在该领域取得了显著的地位。粒子群优化算法(PSO)是最流行的同伦算法,也是研究人员最喜欢的领域之一。在本文中,我们提出了一个简短的概述粒子群算法及其变种解决聚类挑战的适用性。此外,我们提出了一个先进的粒子群算法命名为减法聚类基于边界限制自适应粒子群优化算法(SC-BR-APSO)聚类多维数据。为了比较的目的,我们研究和分析了各种算法,如K-Means,PSO,K-Means-PSO,混合减法+ PSO,BRAPSO,并提出了九个不同的数据集上的算法。提出SC-BR-APSO算法的动机是处理多维数据聚类,具有最小的出错率和最大的收敛速度。

英文摘要:

Data clustering is a recognized data analysis method in data mining whereas K-Means is the well known partitional clustering method, possessing pleasant features. We observed that, K-Means and other partitional clustering techniques suffer from several limitations such as initial cluster centre selection, preknowledge of number of clusters, dead unit problem, multiple cluster membership and premature convergence to local optima. Several optimization methods are proposed in the literature in order to solve clustering limitations, but Swarm Intelligence (SI) has achieved its remarkable position in the concerned area. Particle Swarm Optimization (PSO) is the most popular SI technique and one of the favorite areas of researchers. In this paper, we present a brief overview of PSO and applicability of its variants to solve clustering challenges. Also, we propose an advanced PSO algorithm named as Subtractive Clustering based Boundary Restricted Adaptive Particle Swarm Optimization (SC-BR-APSO) algorithm for clustering multidimensional data. For comparison purpose, we have studied and analyzed various algorithms such as K-Means, PSO, K-Means-PSO, Hybrid Subtractive + PSO, BRAPSO, and proposed algorithm on nine different datasets. The motivation behind proposing SC-BR-APSO algorithm is to deal with multidimensional data clustering, with minimum error rate and maximum convergence rate.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑