一种面向大数据的快速自动聚类算法  

Fast automatic clustering algorithm for big data

作　　者：陈小玉[1] 李晓静[2] 马海英[3] Chen Xiaoyu Li Xiaojing Ma Haiying(College of Computer ＆ Information Engineering, Nanyang Institute of Technology, Nanyang Henan 473000, China Jiyuan Vocational ＆ Technical College, Jiyuan Henan 459000, China College of Computer Science ＆ Technology, Nantong University, Nantong Jiangsu 226019, China)

机构地区：[1]南阳理工学院计算机与信息工程学院,河南南阳473000 [2]济源职业技术学院,河南济源459000 [3]南通大学计算机科学与技术学院,江苏南通226019

出　　处：《计算机应用研究》2017年第9期2651-2654,2658,共5页Application Research of Computers

基　　金：国家自然科学基金资助项目(61402244)

摘　　要：针对大数据环境下高维数据聚类速度慢、准确率低的问题,提出了一种面向大数据的快速自动聚类算法(FACABD)。FACABD聚类算法利用谱聚类算法对大数据集进行归一化和列降维,提出了一种新的快速区域进化的粒子群算法(FRE-PSO),并利用该算法进行行降维;然后在行列降维的基础上,引入聚类模糊隶属度基数,自动发现簇的数目,根据类簇数目,采用FRE-PSO算法结合模糊聚类算法,实现了快速自动聚类。在多个不同数据集上的实验结果表明,该算法能够在数据驱动下快速自动聚类,与其他聚类算法相比,有效地提高了运行速度和精度。Aiming at the problem of slow speed and low accuracy of high dimensional data clustering in big data environment, this paper proposed a fast automatic clustering algorithm for big data （ FACABD）. Firstly, it realized the normalization and row dimension reduction for the large data set by means of spectral Clustering algorithm. Second, it proposed a particle swarm optimization algorithm for fast regional evolution （FRE-PSO） , which could improve the convergence speed and realized the line dimension reduction. And then it introduced the fuzzy cluster membership degree base to automatically discovery the clus- ter number. Finally, it realized fast automatic clustering by the FRE-PSO and fuzzy clustering algorithm. The experiments on multiple, different data sets show that the algorithm can contain the clustering results quickly and automatically by mining data itself, and it can effectively improve the speed and accuracy than the other clustering algorithms.

关 键 词：谱聚类 粒子群优化 区域进化 自动聚类 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TP301.6[自动化与计算机技术—计算机科学与技术]