基于密度分布的K-Means初始聚类中心选择算法

An Improved Algorithm for K-Means Initial Cluster Centers Based on Density Distribution

针对K-Means算法聚类效果的好坏依赖初始聚类中心的选择问题,本文提出一种基于密度分布的简洁K-Means初始聚类中心选择算法.算法利用样本数据相似的稠密程度,较为精准的来寻找初始聚类中心,可有效的克服初始聚类中心选择的盲目性,减少迭代次数及聚类结果的不稳定现象.实验表明,该算法具有良好的聚类效果,稳定性好.

To solve the problem that the choices of K-Means Initial Cluster Centers effect the clustering re-sult, an improved concise algorithm based on density distributionis is proposed. This algorithm finds the initial cluster centers more accurately using the density of sample data distribution,which can effectively avoid the blindness of initial clustering center choice, reduce the number of iterations and the unstable phenomenon of clustering results. The experimental results demonstrate that the algorithm can have a better clustering perform-ance and more stable cluster results.