一种有效的增量聚类算法  

An Effective Incremetal Clustering Algorithm

作　　者：许毕峰[1] 冯少荣[1] 薛永生[1] 刘笑锋[1] 翁伟[1] 

机构地区：[1]厦门大学计算机科学系,厦门361005

出　　处：《计算机科学》2004年第11期93-96,共4页Computer Science

基　　金：福建省自然基金(A0310008);福建省高新技术研究开放计划重点项目资助(2003H043)

摘　　要：聚类是数据挖掘领域中最活跃的研究分支之一,聚类技术在其他的科学领域也有广泛的应用。迄今为止已经提出了大量的聚类算法,其中基于密度的DBSCAN算法因其很多优点而备受关注,为了减少DBSCAN的区域查询次数,降低I/O开销而提出的改进算法有FDBSCAN、LSNCCP等。随着应用的发展,增量聚类显得越来越重要,而现有的增量聚类算法存在很大的局限性。基于LSNCCP,提出了一种有效的增量聚类算法,同时它也可以用于对LSNCCP进行性能优化。Clustering is one of the most flourish direction of data mining. It has been applied abroad at other scientific fields. Many clustering algorithms have been proposed so far,and the DBSCAN algorithm which was density-based was famous for it's advantages. In order to decrease the amount of regional queries and operations of I/O,some people suggested some advanced algorithms such as FDBSCAN,LSNCCP. With the development of application, incremental clustering algorithm became more important,while the incremental clustering algorithms have been suggested have a lot of limitation. Based on LSNCCP, we propose a new effective incremental clustering algorithm called INCCP, which can be used to improve the efficiency of LSNCCP too.

关 键 词：聚类算法 DBSCAN 增量 区域查询 I/O 数据挖掘 性能优化 分支 次数 领域 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论] TP391[自动化与计算机技术—计算机科学与技术]