基于相对密度的混合属性数据增量聚类算法  

Incremental relative density-based clustering algorithm for mixture data sets

作　　者：黄德才[1] 李晓畅[1] 

机构地区：[1]浙江工业大学计算机科学与技术学院,杭州310023

出　　处：《控制与决策》2013年第6期815-822,共8页Control and Decision

基　　金：国家重大科技专项专题项目(2009ZX07318-003-01)

摘　　要：传统的基于密度的带噪声空间数据聚类算法主要存在以下问题:聚类只对具有数值属性的数据有效,而对具有非数值属性的数据失效;参数设置困难且聚类结果对参数较为敏感;聚类的度量以绝对密度值为标准,无法发现密度等级不同的聚类结果.针对以上问题,提出一种面向混合属性数据的、基于相对密度的聚类算法RDBC M,同时提出解决这类问题的增量式聚类算法,并从理论和仿真实验两方面分析、验证了算法的有效性和加速效果.@@@@Traditional density-based clustering algorithm mainly has three problems as follow. Firstly, it only supports spatial attributes without considering non-spatial attributes in the database. Secondly, it is difficult to set the parameters, and the clustering result is sensitive to the parameters. Thirdly, it can’t discover the clusters of different density for adopting absolute density as the metrics of all clusters. In order to overcome these problems mentioned above, the paper presents an relative density-based clustering algorithm for mixture data sets(RDBC M), and further carries out the research on its incremental clustering algorithm. Theoretical analysis and simulation experiment verfy the effectiveness and the performance speed-up effect of the proposed algorithm.

关 键 词：相对密度 绝对密度 混合属性数据集 增量聚类 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]