基于混合密度和微簇聚合的密度峰值聚类算法

Density Peaks Clustering Algorithm Based on Hybrid Density and Micro-clusters Aggregation

密度峰值聚类算法是一种简单高效聚类新算法,但该算法在处理密度分布不均匀数据集时,很难找到正确的类簇中心,并且在样本分配过程中容易出现错误连带现象,导致聚类效果不佳。针对上述问题,提出一种基于混合密度和微簇聚合的密度峰值聚类算法(HMDPC)。HMDPC算法首先根据反向K近邻和样本间的归属关系定义样本的混合密度;其次,将数据划分为多个微簇,定义微簇之间的相似度,基于此相似度对多个微簇进行聚合,从而获得最终的聚类结果。在人工数据集和UCI数据集上进行实验,并将HMDPC算法与其它6种聚类算法比较,实验结果表明HMDPC算法聚类效果较好。

Density peaks clustering algorithm is a simple and efficient new clustering algorithm,but it is difficult to find the correct cluster centers when dealing with datasets with uneven density distribution,and it is easy to appear errors in the process of sample allocation,resulting in poor clustering effect.To solve these problems,a density peaks clustering algorithm based on hybrid density and micro-clusters aggregation(HMDPC)is proposed.The HMDPC algorithm first defines the hybrid density of the samples according to the reverse K nearest neighbor and the attribution relationship between the samples.Secondly,the data is divided into multiple micro-clusters and the similarity between the micro-clusters is defined.Based on this similarity,multiple micro-clusters are aggregated to obtain the final clustering result.Experiments were carried out on synthetic datasets and UCI datasets,and HMDPC algorithm was compared with other 6 clustering algorithms.The experimental results showed that HMDPC algorithm had better clustering effect.