基于遗传模拟退火的K-means聚类方法

针对传统的K-means算法对初始聚类中心的敏感很大,极易陷入局部最优值,基于遗传算法的K-means聚类算法由于个体的多样性不足而常出现早熟等现象,采用遗传模拟退火算法优化初始聚类中心点后进行K-means聚类,并提出了一种新的用于评价聚类结果的适应度函数,该函数更为准确地反映类内距离和类间距离.实验结果表明,该方法能获得更好的聚类结果.

K-means clustering method based on genetic simulated annealing algorithm

As the traditional K-means Clustering Algorithm is excessively sensitive to the choice of the initial cluster centers,it leads to be involved in locally optimal solution.The K-means clustering algorithm usually appears premature phenomenon due to the lack of individual diversity.In order to overcome the above disadvantages,we propose K-means clustering after optimizing the initial cluster center based on genetic simulated annealing algorithm.A new fitness function for evaluation of clustering results is proposed,which is more accurate reflection of the within-class distance and class distance.Experimental results show that the method can get better clustering results.