基于人工鱼群的自适应密度峰值聚类算法
Adaptive density peaks clustering algorithm based on artificial fish swarm
何凯琳 1张正军 1位雅 1唐莉2
作者信息
- 1. 南京理工大学数学与统计学院,江苏南京 210094
- 2. 景德镇学院信息工程学院,江西景德镇 333000
- 折叠
摘要
针对密度峰值聚类算法中截断距离dc和聚类中心缺乏选取依据,以及对簇中存在多密度峰值的数据无法准确聚类问题,提出一种基于人工鱼群的 自适应密度峰值聚类算法(AFSADPC).选择簇中心权值γ大于幂律分布上分位数的样本点作为聚类中心,根据两个相邻簇的簇间边界区域密度与簇平均密度构造簇间合并规则,利用人工鱼群算法寻找使改进轮廓系数指标达到最大值时的最优截断距离dc.在合成数据集和真实数据集上的实验结果表明,AFSADPC算法具有较好的聚类效果.
Abstract
Aiming at the problems that the cutoff distance(dc)and clustering centers are lack of selection basis in density peaks clustering algorithm,as well as the problem that the data with multiple density peaks in a cluster cannot be clustered accurately,an adaptive density peak clustering algorithm based on artificial fish swarm(AFSADPC)was proposed.The sample points whose cluster center weight γ was greater than the quantile on the power-law distribution were selected as the clustering center,and the merging rule between clusters was constructed according to the density of the boundary area between two adjacent clus-ters and the average density of two adjacent clusters,and the artificial fish swarm algorithm was used to find the optimal cutoff distance when the improved silhouette coefficient index reached the maximum value.Experimental results on synthetic datasets and UCI realworld datasets show that AFSADPC algorithm has better clustering effects.
关键词
密度峰值/聚类算法/人工鱼群算法/截断距离/幂律分布/簇合并策略/轮廓系数Key words
density peaks/clustering algorithm/artificial fish swarm algorithm/cutoff distance/power-law distribution/cluster merging strategy/silhouette coefficient引用本文复制引用
出版年
2024
NETL
NSTL
万方数据