基于文化算法的聚类分析  

Cluster analysis based on cultural algorithms

作　　者：刘纯青[1] 杨莘元[1] 张颖[2] 

机构地区：[1]哈尔滨工程大学信息与通信工程学院,黑龙江哈尔滨150001 [2]哈尔滨理工大学经济管理学院,黑龙江哈尔滨150080

出　　处：《计算机应用》2006年第12期2953-2955,2960,共4页journal of Computer Applications

摘　　要：分析了K-均值聚类算法所存在的不足,提出了基于文化算法的新聚类算法,并给出该算法的两个实现版本:CA-version1利用规范知识调整变量变化步长,形势知识调整其变化方向;CA-version2利用规范知识调整变量变化步长及变化方向。文化算法所具有的双层结构特性,使其在问题求解过程中能够利用经验知识来指导搜索过程,从而具有较好的全局寻优性能。仿真实验亦表明,两个版本的文化算法均能有效地克服传统的K-均值算法的缺点,而且全局收敛性能优于基于遗传算法的K-均值聚类算法,同时还可以看出第二个版本的文化算法更适于求解聚类问题。After analyzing the disadvantages of the classical K-means clustering algorithm, a new clustering algorithm based on cultural algorithms was proposed, and two different versions of implementations named CA-versionl and CA-version2 were put forward. CA-version1 uses situational knowledge to control the direction of mutation, and uses normative knowledge to control the step size of mutation. CA-version2 uses normative knowledge to control the step size and the direction of mutation. Cultural algorithms are dual inheritance systems which are different from the others. Because of this feature, the search process is guided by using knowledge abstained from the process of solving problem, which can produce substantial performance improvements. Compared with the classical K-means clustering algorithm, the algorithms based on cultural algorithms, proved by the experimental results, can not only avoid the disadvantages of the classical K-means clustering algorithm, but also have greater searching capability globally than genetic clustering algorithm. Besides, it shows that CA- version2 is more suitable than CA-versionl for clustering problem.

关 键 词：聚类分析 文化算法 K均值算法 信仰空间 规范知识 形势知识 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]