Abstract: In order to improve the efficiency of clustering algorithm, this paper proposes a model of fast closed tree paralleled algorithm on the platform of dynamic cloud. Aiming at the random allocation strategy of cloud computing platform Hadoop, the paper puts forward CDA-GA to meet the requirements of the minimized consumption cost. Moreover, on the foundation of CDA-GA, it proposes the dynamic cloud platform model. The parallelization of traditional frequency closed tree mining algorithm and clustering algorithm and is applied in the dynamic cloud platform, this paper designs a closed tree clustering algorithm framework. Experimental results show that the algorithm is feasible and fits into clustering analysis under massive amounts of data.

Key words: data mining, cloud computing, parallel computing, closed tree, tree clustering, mass data

摘要： 为提高聚类算法效率，提出一种基于动态云平台的快速闭树聚类并行算法。针对云计算平台Hadoop中任务的随机分配策略，给出一个满足最小化消耗成本的任务分配算法CDA-GA，并基于该算法提出动态云平台模型。将传统的频繁闭树挖掘算法与聚类算法并行化，应用于动态云平台中，设计基于动态云平台的闭树聚类算法框架。实验结果表明，该算法有效可行，适合在大规模数据下进行聚类分析。

关键词: 数据挖掘, 云计算, 并行计算, 闭树, 树聚类, 海量数据

CLC Number: 

• TP393