一种高效的K-medoids聚类算法

针对K-medoids算法初始中心点选择敏感、大数据集聚类应用中性能低下等缺点，提出一个基于初始中心微调与增量中心候选集的改进K-medoids算法。新算法以微调方式优化初始中心，以中心候选集逐步扩展的方式来降低中心轮换的计算复杂性。实验结果表明，相对于传统的K-medoids算法，新算法可以提高聚类质量，有效缩短计算时间。

Efficient K-medoids clustering algorithm

Due to the disadvantages of sensitivity to the initial selection of the medoids and poor performance in large data set processing in the K-medoids clustering algorithm, this paper proposed an improved K-medoids algorithm based on a fine-tuned of initial medoids and an incremental candidate set of medoids. The proposed algorithm optimized initial medoids by fine-tu-ning and reduced computational complexity of medoids substitution through expanding medoids candidate set gradually. Expenrimental results demonstrate the effectiveness of this algorithm,which can improve clustering quality and significantly shorten the time in calculation compared with the traditional K-medoids algorithm.