一种基于蜂群原理的划分聚类算法  

Partition clustering algorithm based on artificial bee colony principal

作　　者：刘雷[1] 王洪国[1,2] 邵增珍[2,3] 尹会娟[1] 

机构地区：[1]山东师范大学管理与经济学院,济南250014 [2]山东师范大学信息科学与工程学院,济南250014 [3]山东省分布式计算机软件新技术重点实验室,济南250014

出　　处：《计算机应用研究》2011年第5期1699-1702,共4页Application Research of Computers

基　　金：山东省科技攻关项目(2009GG10001008);济南市高校院所自主创新项目(200906001);山东省软科学研究计划项目(2009RKA285)

摘　　要：针对现有的大部分划分聚类算法受聚类簇的个数K的限制,提出一种基于蜂群原理的划分聚类算法。该方法通过引入蜂群采蜜机制,将聚类中心视为食物源,通过寻找食物源的自组织过程来实现数据对象的聚集。在聚类的过程中引入紧密度函数来评价聚类中心(局部),引入分离度函数来确定最佳聚类簇的个数(全局)。与传统的划分聚类算法相比,本算法无须指定聚类个数即可实现聚类过程。通过仿真实验表明,提出的算法不但对最佳聚类数有良好的搜索能力,而且有较高的准确率:算法时间复杂度仅为O(n×k3)(k<<n),具有较高的执行效率。According to the drawback that most of these algorithms had the shortcoming that clustering results were limited by K value which was the number of clusters,this paper proposed a new partition clustering algorithm based on the principal of artificial bee colony.The clustering method introduced the mechanism of artificial bee colony collecting pollen and every clustering center would be considered as a food source.Then the process of gathering data objects would be achieved by the process of finding the food source.In the process of clustering,proposed tightness function as the fitness to evaluate the cluster center（local） and itroduced separation function to determine the optimal number of clusters（global）.Comapared to traditional partition clustering algorithms,this algorithm did not need the value K that was a given number of clusters to realize clustering process.Simulation results show that the algorithm not only can determine the best number of clusters,and can get a higher clustering accuracy.Furthermore,the time complexity of this algorithm is O（n×k3）（k n）,which is with high efficiency in the implementation.

关 键 词：聚类 划分聚类 人工蜂群 紧密度 分离度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP301.6[自动化与计算机技术—计算机科学与技术]