文献类型：期刊文献

中文题名：基于MIP和改进模糊K-Means算法的大数据聚类设计

英文题名：Design for Large Data Clustering Based on MIP and Improved Fuzzy K-Means Algorithm

作者：陈思慧[1]

机构：[1]广东海洋大学网络与教育技术中心,广东湛江524088

年份：2014

卷号：22

期号：4

起止页码：1270

中文期刊名：计算机测量与控制

外文期刊名：Computer Measurement &Control

收录：CSTPCD、、北大核心2011、北大核心

语种：中文

中文关键词：模糊K均值;聚类;大数据;距离

外文关键词：fuzzy K--Means; clustering; large data; distance

中文摘要：为了克服经典模糊K-Means算法在面对大数据聚类时所出现的聚类效率低和运行时间长的问题,提出了一种基于层次式MPI并行编程模型和改进模糊K-Means算法的大数据聚类方法;首先,引入多层MasterNode节点设计了一种改进的层次式MPI并行编程模型,然后,引入类间距离和类内距离得到一种最优聚类数的计算方式,并设计了一种改进的模糊K-均值聚类算法;采用SlaveNode节点并行运行改进的模糊K均值算法进行数据子集聚类,然后再通过各层MasterNode节点进行汇总和进一步处理;仿真实验表明文中方法能较为精确地实现大数据聚类,准确精确度较经典模糊K均值算法平均约高5.6%,弥补了经典模糊K-Means方法在处理大数据时的正确率低和低效的缺点,具有很强的优越性。

外文摘要：In order to conquer the defects of low clustering efficiency and long operation time in the classic fuzzy K--Means algorism, an improved algorism based on MPI （Message Passing Interface） and improved fuzzy K--Means algorithm was proposed. Firstly, the improved hierarchical MPI model was introduced by add the multiple--hierarchical MasterNodes, then a optimal duster number computing method was obtained by adding the distance in cluster and out of the cluster, the improved fuzzy K--Means algorithm was also obtained. The SlaveNode was used to operate the improved fuzzy K--Means algorithm in currency to cluster multiple sub--data sets, then it transfer the clustering re- sult to MasterNode to be further managed. The simulation experiment shows the method in this paper can realize the large data clustering ac- curately, the precision was improved by 5.6% compared with classic fuzzy K--Means algorism, so it conquers the low correctness rate and efficiency, therefore, it is proved to be having big priority.