一种基于变异蝙蝠算法的高维聚类方法  

High-dimensional Clustering Method Based on Variant Bat Algorithm

作　　者：寇广 汤光明 何嘉婧 张恒巍 Kou Guang;Tang Guangming;He Jiajing;Zhang Hengwei(The PLA Information Engineering University, Zhengzhou 450001, China;Science and Technology on Information Assurance Laboratory, Beijing 100072, China)

机构地区：[1]中国人民解放军信息大学,郑州450001 [2]信息保障技术重点实验室,北京100072

出　　处：《系统仿真学报》2018年第4期1253-1259,共7页Journal of System Simulation

基　　金：国家自然科学基金(61303074),信息保障技术重点实验室开放基金(KJ-14-106)

摘　　要：随着大数据时代的来临,信息资源迅猛增长,数据逐渐趋于高维化。传统的聚类方法针对低维数据有较好的效果,而不再适用于高维数据。在目前已有的高维聚类算法的基础上,提出一种基于智能优化算法的高维聚类算法SSC-BA(Soft Subspace Clustering based on Bat Algorithm)。算法设计了一个新目标函数,结合了加权类内相似性及类间差异性和界约束权值矩阵,引进了一种变异蝙蝠算法计算权值矩阵,给出了新的学习规则。对提出的算法进行了仿真实验,与其他软子空间聚类算法进行对比测试。实验结果表明此聚类算法适用于高维数据并较其它算法有一定的性能优势。With the advent of the era of big data, the information resource is growing rapidly, and the data are becoming high-dimensional. Traditional clustering methods have a good effect for low-dimensional data, but no longer apply to high-dimensional data. On the basis of existing high-dimensional clustering algorithm, a high-dimensional clustering algorithm based on intelligent optimization SSC-BA is proposed. A novel objective function is designed, which integrates the fuzzy weighting within-cluster compactness and the between-cluster separation. A variant bat algorithm is introduced to calculate the weight matrix, giving the new learning rules. Simulation experiments are made for the proposed algorithm, and other soft subspace clustering algorithm is compared with the test. Experimental results show that the clustering algorithm is suitable for high-dimensional data, and has certain performance advantages compared with other algorithms.

关 键 词：高维聚类 权值矩阵 蝙蝠算法 变异策略 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]