可变网格优化的K-means聚类方法  

K-means Clustering Method Based on Variable Grid Optimization

作　　者：万静[1] 张超[1] 何云斌[1] 李松[1] 

机构地区：[1]哈尔滨理工大学计算机科学与技术学院,哈尔滨150080

出　　处：《小型微型计算机系统》2018年第1期95-99,共5页Journal of Chinese Computer Systems

基　　金：黑龙江省教育厅科学技术研究项目(12531z004)资助

摘　　要：传统k-means算法需要人为指定聚类数k,对初始中心点的选取比较敏感,只能发现球状簇.针对k-means算法的不足,提出了基于可变网格优化的k-means聚类算法,该算法通过可变网格划分解决了随机选取初始中心点不具有代表性的问题,同时排除了噪声的干扰.此外,针对最大密度不唯一的情况进行了研究,选取各距离最大的类簇为最优类簇.最后,基于可变网格优化的k-means算法结合BWP指标对最佳聚类数进行了优化,解决了最佳聚类数事先无法确定的问题.理论和实验结果表明,基于可变网格优化的k-means算法具有更好的有效性和可行性.The traditional k-means algorithm needs to specify the number of clusters of k,the initial center of the selection is relativelysensitive, moreover, the k-means algorithm can only be found in spherical clusters. In view of the above shortcomings of k-means algo-rithm, this paper proposes a k-means clustering algorithm based on variable grid optimization, the algorithm solves the random selec-tion of initial centers are not representative of the problem by variable grid,and eliminate noise interference. In addition,for the casewhere the maximum density is not unique, the cluster with the largest distance is selected as the best cluster. Finally, the k-means algo-rithm based on variable grid optimization is combined with the BWP index to optimize the optimal clustering number, which solves theproblem that the optimal clustering number can not be determined beforehand. The combination of theory and experiment shows thatthe k-means algorithm based on variable mesh optimization has better effectiveness and feasibility.

关 键 词：K-MEANS聚类算法 可变网格 初始中心点 BWP指标 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]