K-means算法的初始点优化研究

为了克服经典K-means算法对初始聚类中心过分依赖的缺点,该文提出采用竞争神经网络和密度思想对经典k-means算法进行预处理,从而改变经典K-means算法对初始聚类中心的随机选择。实验结果表明,这两种方法是有效的。

Study on the Initial Centrists of K-means Algorithm

In order to conquer the problem that k-means algorithm depends on initial cluster centrists,so this paper discusses use competition neural network and the mind of density to improve the classic k-means algorithm.The two methods are able to improve the random choice of the initial centrists in the classic k-means algorithm.Experimental results show that the two algorithms are effective.