基于MapReduce的网格化优化CURE算法的实现  

Gridding Optimization CURE Algorithm Implemented based on MapReduce

作　　者：郎福通 王鹏[1] 

机构地区：[1]成都信息工程学院软件工程学院,四川成都610255

出　　处：《成都信息工程学院学报》2014年第6期603-608,共6页Journal of Chengdu University of Information Technology

基　　金：四川省青年科学基金资助项目(09ZQ026-068)

摘　　要：针对CURE算法处理大量数据时聚类速度较慢的问题,一方面采用网格聚类方法对初始聚类对象进行网格预聚类处理,缩短初始化族聚类时间;另一方面采用MapReduce框架对算法进行并行性扩展,使其能够充分利用集群的计算和存储能力,从而加速海量数据的处理。以联合程序开发网站的数据集和MATLAB人工数据集作为测试数据集,对改进算法Grid-CURE进行实验分析。实验结果表明:方法可有效提升处理大数据的效率以及提升其抗噪声能力。As handling large amounts of data,the efficiency of the CURE algorithm is not very well. For the sake of solving this problem,on the one hand,gridding clustering is used to process the raw data. On the other hand the MapReduce computing frame which makes full use of computing and storage capacity to accommodate the processing of massive data is chosen to parallelize the CURE algorithm. To analyze the effect of the improved Grid-CURE algorithm,tow data sets are prepared. One of them is artificial data set which is created by MATLAB tool. The other comes from programmers united develop net. Experimental results show that the improved method could effectively accelerate the efficiency of the cure algorithm and enhance its anti-noise ability.

关 键 词：CURE算法 网格聚类 MAPREDUCE 分布式聚类 Grid-CURE算法 

分 类 号：TP311.1[自动化与计算机技术—计算机软件与理论]