基于模拟退火算法对K-means聚类算法的优化  

Optimization for K-means Clustering Algorithm Based on Simulated Annealing Algorithm

作　　者：刘寒梅[1] 张鹏[1] 

机构地区：[1]长春工业大学计算机科学与信息工程学院,吉林长春130012

出　　处：《中国西部科技》2013年第6期23-24,71,共3页Science and Technology of West China

摘　　要：K-means聚类算法是近年来数据挖掘学科的一个研究热点和重点,该算法是基于划分的聚类分析算法.目前这种算法在聚类分析中得到了广泛应用。本文将介绍K-means聚类算法的主要思想,及其优缺点。针对该算法经常陷入局部最优,以及对孤立点敏感等缺点,提出了一种基于模拟退火算法的方法对其进行优化,可以有效地防止该算法陷入局部最优的情况。K-means clustering algorithm, which is a division-based clustering and analysis algorithm, has become a hotspot of data-mining subject in recent years. Now this algorithm has been widely applied in the clustering analysis. In this article, we introduced the main idea and advantages/disadvantages of the K-means clustering algorithm. Aiming to the defects of this algorithm such as local optimum and sensitive to isolated points, we suggested a simulation-based annealing algorithm to optimize it so as to prevent the algorithm from experiencing local optimum efficiently.

关 键 词：数据挖掘 聚类算法 K-MEANS聚类算法 模拟退火算法 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]