一种 K-均值聚类的改进算法及其应用

摘要

由于传统的K-均值聚类算法固有的特性，如对初始聚类中心的依赖性和对噪声点的敏感性，导致了其聚类结果的不稳定。文中基于 k-dist图提出一种改进算法，算法首先去除数据集中的噪声点，然后从数据集中选取靠近点聚集区域、相距最远的 k′个数据点作为初始簇中心。实验结果证明，文中算法能够很好地消除K-均值聚类算法对初始簇中心的依赖性，并能有效去除噪声点。%Because of the inherent characteristics of the traditional K‐means algorithm ,such as the dependence of the initial clustering center and the sensitivity to noise points ,it is easy to generate the instable clustering results .An improved algorithm is proposed based on the k‐dist graph .The algorithm first removes the noise points of the data set ,and then selects k′initial center ,w hich is close to the point gathering area and far away from each other .The experimental results show that the algorithm can eliminate its dependence of the initial center as well as removing the noise points effectively .