局部网格动态聚类算法  

Dynamic clustering algorithm based on local grid

作　　者：王玮琪 万仁霞 周方祥 WANG Weiqi;WAN Renxia;ZHOU Fangxiang(School of Computer and Engineering,North Minzu University,Yinchuan 750021,China)

机构地区：[1]北方民族大学计算机科学与工程学院

出　　处：《现代电子技术》2020年第1期102-106,共5页Modern Electronics Technique

基　　金：国家自然科学基金资助项目（61662001）;国家自然科学基金资助项目（61762002）;国家民委领军人才支持计划资助项目（2016GQR06）

摘　　要：针对传统网格聚类算法聚类精度较低,处理流数据效率较低等问题进行改进。提出局部网格动态聚类算法,算法引入维度半径概念进行增量动态网格划分,通过采用新的簇边界判定方法对簇边界进行判定,依据稀疏网格与其邻接密集网格的质心距离,将稀疏网格归并到相应网格簇中,对于不能归并的稀疏网格则采用局部网格划分方法对稀疏网格再次进行划分聚类,避免簇边界的误删,在一定程度上提高了聚类精确度。通过对比实验结果表明提出的算法具有更好的聚类时效性和聚类精度。A dynamic clustering algorithm based on local grid is proposed to deal with the facts of low clustering accuracy and low efficiency in processing streaming data in the traditional grid clustering algorithm.In this algorithm,the concept of dimension radius is introduced for incremental dynamic grid division,and a new judgment method of cluster boundary is adopted to determine the boundary of clusters.In addition,the sparse grids are merged into the corresponding grid clusters according to the centroid distances from sparse grids to their neighboring dense grids.As for those who cannot be merged,the local grid division method is adopted to divide and cluster them again,so as to avoid mistaken deletion of cluster boundaries and improve the clustering accuracy to a certain extent.The comparative experiment results show that the proposed algorithm has better clustering timeliness and accuracy.

关 键 词：网格 局部密度 聚类算法 密集网格 稀疏网格 簇边界 

分 类 号：TN911.1-34[电子电信—通信与信息系统] TP311[电子电信—信息与通信工程]