基于引力搜索机制的数据聚类及特征选择算法  

Data clustering and feature selection algorithm based on gravitational search mechanism

作　　者：张雪峰 杜孝平[2] 王晓健 王哲 ZHANG Xue-feng;DU Xiao-ping;WANG Xiao-jian;WANG Zhe(Department of Information Management and Consultation,Beijing Sai Di Industrial and Information Engineering Supervision Center Limited Company,Beijing 100048,China;School of Software,Beihang University,Beijing 100191,China;School of Information Engineering,Xiangtan University,Xiangtan 411105,China)

机构地区：[1]北京赛迪工业和信息化工程监理中心有限公司信息化管理与咨询部,北京100048 [2]北京航空航天大学软件学院,北京100191 [3]湘潭大学信息工程学院,湖南湘潭411105

出　　处：《计算机工程与设计》2021年第9期2536-2544,共9页Computer Engineering and Design

基　　金：湖南省自然科学基金项目(2018JT1025)。

摘　　要：为同步选择具有相关特征的数据聚类数量,提出一种基于引力搜索机制的聚类和特征选择算法。设计一种代理表示策略实现聚类中心和特征数量的编码;提出一种动态临界值方法决定聚类和特征数量,通过代理适应度的不断评估寻找最优聚类量和相关特征;分析算法的时间复杂度,通过8个经典数据集测试算法性能,并与7种常规数据聚类算法作对比。实验结果表明,所提算法在聚类和特征数量选择上具有更高的准确率。To synchronously select the number of data clusters along with relevant features,a clustering and feature selection algorithm based on gravitational search mechanism was proposed.An agent representation strategy was designed to encode the clustering center and the number of features.A dynamic threshold method was proposed to determine the clustering and the number of features,and the optimal number of clusters and the relevant features were found according to the continuous evaluation of the agent fitness.The time complexity of the proposed algorithm was analyzed,the performance of the algorithm was tested with eight classical data sets and compared with seven conventional data clustering algorithms.Experimental results show that the proposed algorithm has higher accuracy in clustering and feature number selection.

关 键 词：数据聚类 特征选择 引力搜索 适应度 聚类中心 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]