改进的K-means聚类算法及应用

Improved K-means Algorithm and Its Application

传统的K-means算法需要事先确定初始聚类中心,聚类精确程度不高。针对以上问题,本文结合熵值法和动态规划算法来对传统的K-means算法进行改进,提出了基于熵值法及动态规划的改进K-means算法。熵值法用来修订算法的距离计算公式,以提高算法的聚类精确程度,动态规划算法用来确定算法的初始聚类中心。将改进算法应用于矿井监测传感器聚类中,结果显示较传统的K-means算法,改进算法效率有了明显提高,聚类精确程度有较大增强。

The traditional K-means has sensitivity to the initial clustering centers,and its clustering accuracy is low.To against these short comings,an improved K-means algorithm based on the combination of dynamic programming algorithm and entropy method is proposed.The entropy method is used to amend the distance calculating formula to improve the clustering accuracy,and dynamic programming algorithm is used to define the initial cluster centers.The result of the simulation on the clustering in the mine monitoring sensors shows that the proposed algorithm has better performance than the traditional K-means algorithm in terms of efficiency and clustering accuracy.