曾俊

摘  要： 为提升数据挖掘中聚类分析的效果，在分析数据挖掘、聚类分析、传统K?means算法的基础上，提出一种改进的K?means算法。首先将整体数据集分为[k]类，然后设定一个密度参数为[?]，该密度参数反映数据库中数据所处区域的密度大小，[?]值与密度大小成正比，通过密度参数优化[k]个样本数据的聚类中心点选取;依据欧几里得距离公式对未选取的其他数据到各个聚类中心之间的距离进行计算，同时以此距离为判别标准，对各个数据进行种类划分，从而得到初始的聚类分布;初始聚类分布得到之后，对每一个分布簇进行再一次的中心点计算，并判断与之前所取中心点是否相同，直到其聚类收敛达到最优效果。最后通过葡萄酒数据集对改进算法进行验证分析，改进算法比传统K?means算法的聚类效果更优，能够更好地在数据挖掘当中进行聚类。

关键词： 数据挖掘; 聚类分析; K?means聚类算法; 聚类中心选取; K?means算法改进; 初始中心点

中图分类号： TN911.1?34                       文献标识码： A                          文章编号： 1004?373X（2020）03?0014?04

Analysis of partition?based data mining K?means clustering algorithm

ZENG Jun

（College of Big Data and Intelligent Engineering， Yangtze Normal University， Chongqing 408100， China）

Abstract： An improved K?means algorithm on the basis of the analysis of data mining， clustering analysis and traditional K?means algorithm is proposed to improve the effect of clustering analysis in data mining. The whole data set is divided into [k] classes firstly， and then a density parameter [?] is set， which reflects the density of the area in which the data is located in the database. The value of [?] is proportional to the density. The selection of cluster center points of [k] sample data is optimized by the density parameter. The distance between other unselected data and each cluster center is calculated by Euclidean distance formula. At the same time， the distance is taken as the criterion to divide each data into different categories， so as to get the initial clustering distribution. After the initial clustering distribution is obtained， the center point of each distribution cluster is calculated again to judge whether it is the same as the previous center point， until the clustering convergence reaches the optimal effect. Finally， the wine algorithm is used to verify and analyze the improved algorithm. The results show that the improved algorithm has better clustering effect than the traditional K?means algorithm， and can better perform clustering in data mining.

Keywords： data mining; clustering analysis; K?means clustering algorithm; clustering center selection; K?means algorithm improvement; initial center point

0  引  言

信息时代的来临及其快速发展为人们的生活与工作带来了极大便利，数据资源对于人们的影响也越来越大。如何从庞大繁杂的数据资源当中汲取有效信息变得极其重要，数据挖掘应用而生。聚类分析在数据挖掘模型当中的作用巨大。基于划分的K?means算法是聚类分析中具有代表性的算法，其收敛性相对较强，但传统的K?means仍存在诸多问题，对于K?means算法应当如何改进，怎样通过算法融合和各类技术手段进行优化，很多专家学者对此相当关注。……