一种改进的K_means聚类方法

An Improved K_means Clustering Algorithm

针对传统K_means聚类方法采用随机选择初始聚类中心而导致的收敛速度慢的问题,本文结合空间中的距离度量提出一种改进的K_means聚类算法。该方法通过给出有效的启发式信息,选择较好的聚类中心,减少聚类达到稳定状态所需要的迭代步骤,加速算法的执行。标准数据集上的实验结果表明,与传统的K_means聚类方法相比,本文提出的改进的聚类方法收敛速度快,从而在较少的迭代后得到良好的聚类效果。

This paper presents an improved clustering model based on distance measurement, in order to solve the problem of slow convergence rate of traditional Kmeans clustering method by selecting initial cluster centers randomly. By using effective heuristic information, this method selects better clustering centers and reduces the iteration steps of attaining stable clustering state. Then the speed of algorithm is accelerated. Simulation results on UCI datasets demonstrate that comparing with traditional K_ means clustering means, the improved K_means has fast convergence rate and the better clustering results are obtained by this model after less iterations.