一种新的基于图划分聚类算法——GAGPBCUK算法

Novel Graph Partition Based Clustering Algorithm——GAGPBCUK

提出了一种新的基于图划分的聚类算法——GAGPBCUK算法。该算法解决了谱聚类算法参数敏感和聚类结果不准确等问题。3组仿真实验结果表明,GAGPBCUK算法不仅在识别和学习数据集中的隐含聚类数方面具有很好的性能,而且能够得到比谱聚类算法(NJW算法)更加有效的聚类结果。

A novel graph partition based clustering algorithm（GAGPBCUK）was proposed to prevent the defects of spectral clustering methods,such as sensitive parameters and inaccurate results.Experiment results on three simulation datasets indicate that the proposed algorithm can not only determine and learn the dataset’s cluster number effectively but also can get more effective clustering result than spectral clustering algorithm（i.e.NJW algorithm）.