基于闻香识源的改进人工蜂群聚类算法  

An Improved ABC Clustering Algorithm Based on Scent of a Source

作　　者：李梅莲[1] 郭超峰[1] LI Meilian GUO Chaofeng(School of Information Engineering, Xuchang University, Henan Xuchang 461000, China)

机构地区：[1]许昌学院信息工程学院,河南许昌461000

出　　处：《河南大学学报（自然科学版）》2017年第5期552-559,共8页Journal of Henan University:Natural Science

基　　金：河南省科技攻关项目(132101110095;122102210488);河南省教育厅科学技术研究重点项目(13A520748)

摘　　要：为了克服传统K-Means算法初始聚类中心选择的盲目性,提高聚类精度和聚类结果的稳定性,提出一种基于闻香识源的人工蜂群聚类算法,用于数据聚类.该算法首先利用样本数据稠密度反馈的信息(花香)来寻找初始聚类中心,接着交替进行K-Means聚类,人工蜂群在高密度数据区以贪婪原则搜索最佳聚类中心,往复多次以达到良好且稳定的聚类效果.实验表明该算法简单高效,聚类效果好.In order to avoid choosing initial clustering centers of k-means algorithm blindly and improve the precision and stability of the clustering results,an improved ABC clustering algorithm based on scent of a source is presented.It combines ABC algorithm with k-means algorithm together to be used in data clustering.First of all,this algorithm calculates the number of data points around each data point within a certain radius as each data point＇s density（flower scent）.Next this algorithm searches the initial cluster centers based on these data densities,then uses k-means to cluster,uses artificial bees to search for better cluster centers in high-density data area by greedy,and repeat them many times to achieve more stable and better clustering results.Experiments showed that this algorithm was simple and efficient in data clustering.

关 键 词：蜂群算法 数据聚类 K-Means 初始聚类中心 距离矩阵 密度分布 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]