一种不同的基于数据分区的并行构建密度树聚类算法(PCDTC)  

A Much Different Parallel Construction Density Tree Clustering (PCDTC) Algorithm Based on Data Partitioning

作　　者：张云鹏[1] 张璐[2] 翟正军[1] 马春燕[1] 戴维迪[2] 

机构地区：[1]西北工业大学,陕西西安710072 [2]天津大学,天津300072

出　　处：《西北工业大学学报》2008年第4期524-529,共6页Journal of Northwestern Polytechnical University

基　　金：西北工业大学科技创新基金(W016141)资助

摘　　要：传统的聚类算法在全空间下的聚类过程倾向于输出单一的聚类结果,高维数据在不同的子空间多视图下往往呈现不同的数据结构。文中引入空间的正交化方法实现在不同子空间上的并行化,构建密度树聚类,以提供对数据集在多维子空间视图下聚类结果的多样性观测,通过F-measure值引导用户确定不同子空间中感兴趣的聚类结果。真实数据集上的实现结果证明了上述方法的有效性。Aim. Like Ref. 7 by Xia et al, we deal with the subject stated in the title; unlike Ref. 7,our method is much different and, we believe, more efficient. In the full paper, we explain our algorithm and its performance in some detail; in this abstract, we just add some pertinent remarks to listing the three topics of explanation. The first topic is. the essentials of our algorithm. The second topic is. PCDTC algorithm. In this topic, we use our algorithm to partition high-dimensional data into orthogonal subspaces as shown in Fig. 2. The third topic is. the analysis of the performance of PCDTC algorithm. The clustering results and their comparison with two other traditional methods, given in Tables 1 and 2 in the full paper, show preliminarily that our algorithm not only provides various types of effective clustering results but also enhances the clustering efficiency with adequate clustering precision ensured.

关 键 词：并行处理 数据处理 聚类 分区 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]