一种K-均值优化算法的研究与改进  

Research and Improvement of K-means Optimization Algorithm

作　　者：孙艺 赵瑛珲 王天棋 马彦凯 赵佳琪 SUN Yi;ZHAO Ying-hui;WANG Tian-qi;MA Yan-kai;ZHAO Jia-qi(School of Computer Science,Beijing University of Posts and Telecommunications,Beijing 100876 China;International School,Beijing University of Posts and Telecommunications,Beijing 100876 China;School of Economics and Management,Beijing University of Posts and Telecommunications,Beijing 100876 China;Tech Development Department,Shanghai Zhiwei Robot Co.,Ltd.,Shanghai 200120 China;Label Evaluation Department,China National Institute of Standardization,Beijing 100191 China)

机构地区：[1]北京邮电大学,计算机学院,北京100876 [2]北京邮电大学,国际学院,北京100876 [3]北京邮电大学,经济管理学院,北京100876 [4]上海智位机器人股份有限公司,开发部,上海200120 [5]中国标准化研究院,标准评估部,北京100191

出　　处：《自动化技术与应用》2021年第9期1-5,11,共6页Techniques of Automation and Applications

基　　金：河北省重点研发计划项目(编号20313701D);河北省重点研发计划项目(编号19210404D)。

摘　　要：传统K-均值聚类算法处理数据效率低下,而且结果偏差较大。为此,本文涉及一种优化算法,通过衡量处罚方式的程度控制算法迭代方式,以计算所得簇的平均误差的数值为依据,计算簇分配权值的大小,再用加权准则函数计算簇集中的加权距离,将取值最小的簇作为样本点,筛选掉平均误差较大的簇,从而提高算法的效率。用本文设计的算法与传统K-均值算法相比较,以含有大量噪音的数据集为实验数据,发现在抗噪性、聚类效果和运行稳定性方面,本文算法都明显优于传统算法。The traditional K-mean clustering algorithm is low efficiency in processing data,and leads to the large deviation of the results.For this,this paper involves an optimization algorithm,by measuring the degree of punishment way control algorithm itera-tively,to calculate the average error on the cluster,on the basis of the numerical computing the size of the cluster distribution value,then the cluster is calculated by the weighted criterion function weighted distance,the minimum value of cluster as a sample point,average error filter out larger clusters,so as to improve the efficiency of the algorithm.Comparing the algorithm designed in this paper with the traditional K-mean algorithm and taking the data set containing a lot of noise as experimental data,it is found that the algorithm in this paper is significantly better than the traditional algorithm in terms of noise resis-tance,clustering effect and operation stability.

关 键 词：K-均值算法 迭代方式 加权准则函数 粗糙集 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]