基于改进的密度峰值算法的K-means算法  

K-means Algorithm Based on Improved Density Peak Algorithm

作　　者：杜洪波[1] 白阿珍 朱立军[2] Du Hongbo;Bai Azhen;Zhu Lijun(School of Science,Shenyang University of Technology,Shenyang 110870,China;School of Information and Computation Sciences,North University for Nationalities,Yinchuan 750021,China)

机构地区：[1]沈阳工业大学理学院,沈阳110870 [2]北方民族大学信息与计算科学学院,银川750021

出　　处：《统计与决策》2018年第18期20-24,共5页Statistics & Decision

基　　金：国家自然科学基金资助项目(61362033)

摘　　要：针对传统K-means算法存在的随机选取初始聚类中心和类簇数目需要人为选定,从而导致聚类结果不稳定,容易陷入局部最优解的问题,文章提出了一种基于改进的密度峰值算法（DPC）的K-means算法,该算法首先采用改进的DPC算法来选取初始聚类中心,弥补了K-means算法初始聚类中心随机选取导致易陷入局部最优解的缺陷;其次运用K-means算法进行迭代,并且引入熵值法计算距离优化聚类。在UCI数据集上的实验表明,该算法得到较好的初始聚类中心和较稳定的聚类结果,并且收敛速度也较快,证明了该算法的可行性。The initial clustering centers and the number of clusters need to be selected manually in traditional K-means algorithm, so the resuh of clustering is unstable and easy to fall into local optimal solution. To deal with this problem, this paper proposes a K-means algorithm based on the improved algorithm of density peak （DPC）. The proposed algorithm firstly uses the improved DPC algorithm to select the initial （;lustering center, so as to make up for the flaw that the random selection of initial clustering center of [k-means] algorithm leads to the easily trapped local optimal solution, and then uses the K-means algorithm to ilerate and introduee the entropy method to calculate the distance to optimize clustering. The result of experiment on the UCI dataset shows that the proposed algorithm can obtain relatively better initial clustering centers and relatively more stable clustering resuits, with a faster convergence, thus proving the feasibility of the algorithm.

关 键 词：K-MEANS算法 改进的DPC算法 聚类 熵值法 初始聚类中心 优化聚类 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]