基于距离和密度的d-K-means算法.PDF

37 6 Vol. 37 No. 6 第 卷第 期 计算机应用研究 录用定稿 Application Research of Computers Accepted Paper 基于距离和密度的d-K-means 算法 * 唐泽坤，朱泽宇，杨 裔，李彩虹，李 廉 (兰州大学 信息科学与工程学院, 兰州 730000) 摘 要：K-means 算法有实现简单、速度快的特点，是应用最广泛的聚类算法。针对K-means 算法对初始聚类中心 和噪声敏感的缺点，提出了d-K-means 算法（distancedensity），在K-means 算法的基础上权衡了密度和距离对聚 类的影响，对数据进行加权处理，在权值基础上引入“最小最大原则”选择初始聚类中心，自动确定类中心个数。 实验结果表明，d-K-means 算法在低维数据与高维数据上都可以取得较好的聚类效果，并且更好地应对低密度区域 数据，更好地进行类中心选择。 关键词：聚类；K-means 算法；最小最大原则；类中心个数 中图分类号：TP301.6 doi: 10.19734/j.issn.1001-3695.2018.10.0861 D-K-means algorithm based on distance and density Tang Zekun, Zhu Zeyu, Yang Yi, Li Caihong, Li Lian (College of Information Science Engineering , Lanzhou University, Lanzhou 730000, China) Abstract: K-means algorithm has the characteristics of simple implementation and fast speed, and is the most widely used clustering algorithm. To deal with the disadvantages of the K-means algorithm which is sensitive to initial clustering center and noise, the d-K-means algorithm is proposed. On the basis of the K-means algorithm, the d-K-means algorithm weighs the impact of density and distance on clustering to weight the data, and selects the initial clustering center by introducing the minimax principle on the basis of weight, and automatically determines the number of class centers. Experimental results show that d-K-means algorithm can achieve better clustering results on low-dimensional data sets and high-dimensional data sets, and better deal with low-density regional data , and better select class centers