基于流数据的模糊聚类算法  

FUZZY CLUSTERING ALGORITHM BASED ON STREAMING DATA

作　　者：龚静[1] 王翰虎[2] 

机构地区：[1]铜仁学院计算机科学系,贵州铜仁554300 [2]贵州大学计算机科学系,贵州贵阳550025

出　　处：《计算机应用与软件》2008年第2期250-252,277,共4页Computer Applications and Software

摘　　要：对流数据进行有效聚类是一个吸引研究者很大注意力的问题。传统的聚类挖掘算法只能适用于纯数值属性数据或纯分类属性数据,很难适用于混合属性的数据。针对混合属性数据的特点,在借鉴AcluStream算法的基础上,提出了一种模糊聚类算法。算法对流数据的相异度分类度量,定量属性使用欧氏距离和曼哈坦距离度量,定性属性可以采用hamming距离度量。模糊聚类算法的主要步骤有两步:第一步,运用最小距离聚类算法进行聚类,构成一个初始类。第二步,对基于最小距离聚类算法进行聚类所得到的初始簇,运用密度聚类方法进行聚合或分割,使得聚类集合稳定。实践证明:该算法是快速地有效的。Effective clustering of streaming data is attractive to researchers. The traditional clustering excavation algorithm is only suitable for the data of cardinal number value attribute or the data of pure classified attribute, and it is not suitable for data of mixed attribute. Considering the characteristic of the data of mixed attribute, a fuzzy clustering algorithm is presented based on the AcluStream algorithm. The differentiation of the streaming data is classified and measured. The quota attribute is measured by Euclidean space ＆ Kazak Tanzania, and the qualitative attribute is measured by hamming. The fuzzy clustering algorithm has two steps. Firstly, the minimum clustering algorithm is applied to constitute an initial cluster. Secondly ,the initial cluster is clustered and segmented by density clustering to make the cluster set stable. The practice proves that the algorithm is effective and fast.

关 键 词：流数据 流数据挖掘 模糊聚类算法 

分 类 号：TP274.2[自动化与计算机技术—检测技术与自动化装置] TP311[自动化与计算机技术—控制科学与工程]