新型群智能算法在求解聚类问题上的对比研究  

Comparative Study on Several New Swarm Intelligence Algorithms in Solving Clustering Problems

作　　者：薛锋[1,2,3] 刘泳博 陈逸飞 XUE Feng;LIU Yong-bo;CHEN Yi-fei(School of Transportation and Logistics,Southwest Jiao Tong University,Chengdu Sichuan 611756,China;Graduate school of Tangshan,Southwest Jiao Tong University,Southwest Jiao tong University,Tangshan Hebei 063000,China;National Engineering Laboratory of Integrated Transportation Big Data Application Technology,Southwest Jiao Tong University,Chengdu 611756,China)

机构地区：[1]西南交通大学交通运输与物流学院,四川成都611756 [2]西南交通大学唐山研究生院,河北唐山063000 [3]西南交通大学综合交通大数据应用技术国家工程实验室,四川成都611756

出　　处：《计算机仿真》2023年第3期370-376,共7页Computer Simulation

基　　金：国家自然科学基金项目(61203175);四川省科技计划项目(2019YJ0211);综合交通大数据应用技术国家工程实验室开放基金项目(CTBDAT201902,CTBDAT201911)。

摘　　要：为更好地解决kmeans聚类算法对初始聚类中心敏感的问题,选取了最近几年比较典型的5种群智能算法,如蝙蝠算法、磷虾算法、灰狼算法、鲸鱼算法和麻雀算法等,对Kmeans算法在初始聚类中心的选择上进行优化,然后通过6个UCI中典型的数据集进行聚类仿真,并分别从算法的求解精度、稳定性、收敛速度和聚类性能等方面对比分析了各算法的性能。仿真结果表明,灰狼算法在求解聚类问题时效果最佳,能够在较短的时间内有效地对聚类中心进行优化,并且具有较高的精度和稳定性,可有效地解决kmeans算法对初始聚类中心敏感的问题。In order to better solve the problem that the Kmeans clustering algorithm is sensitive to the initial clustering center,five typical intelligent algorithms of the population in recent years were selected to.optimize the Kmeans algorithm in the selection of the initial clustering center,such as the bat algorithm,the krill algorithm,the gray wolf algorithm,the whale algorithm,and the sparrow algorithm.Then the performances of each algorithm were compared and analyzed from the aspects of the algorithm's solution accuracy,stability,convergence speed and clustering performance by the clustering simulation experiments which were carried out through 6 typical data sets in UCI.The simulation results show that the gray wolf algorithm has the best effect when solving the clustering problem;It can effectively optimize the clustering center in a short time,and has high accuracy and stability;It can effectively solve the kmeans algorithm.The initial cluster center is sensitive.

关 键 词：聚类算法 新型群智能算法 聚类中心优化 仿真 性能对比 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]