基于自然和加权共享最近邻的密度峰值聚类算法

打开文本图片集

中图分类号：TP181 文献标志码：A

Density Peak ClusteringAlgorithm Based on Natural and Weighted Shared Nearest Neighbors

Wang Sen， Chen Xiang，Zhan Xiaoqin， Xu Lu， Wu Qizheng （SchoolofScience，EastChinaJiaotongUniversity，Nanchang33oo13，China）

Abstract： Density peak clustering （DPC）has been widely used as an eficient and non-iterativeclustering algorithm.However，studies have found that DPC struggles to select correct cluster centers，especially in datasets with non-spherical clusters and non-uniformdensity.Moreover，DPCis heavily influenced bythe truncation distance parameter.In order to address the issue of poor performance of DPCon datasets with uneven density distributions，adensity peak clustering algorithm based on natural and weighted shared nearest neighbors is proposed. It first introduced natural nearest neighbor computations tocalculate weights.Then，it redefined thesimilaritybetweendata objects basedonthe definitionsoffirst-order and second-order shared nearestneighbors.Subsequently，by fusing the definitions of shared nearest neighbor similarityand natural nearest neighbor weights，relative density and relative distance were calculated.Finaly，anovel strategy for distributing cluster centers Was designed.

Key words：clustering algorithm; density peak clustering; natural nearest neighbor; shared nearest neighbor; cluster center diffusion

Citationformat：WANG S，CEHN X， ZHANX，et al.Density peak clustering algorithm based on natural and weighted shared nearest neighbors[J].Journal ofEast China Jiaotong University，2025，42（4）： 120-126.

聚类是机器学习中的一种分类方法，旨在根据数据集的某些特征将其划分为多个潜在的簇。（剩余8680字）