一种基于共享近邻的密度聚类算法  

A DENSITY CLUSTERING ALGORITHM BASED ON SHARED NEAREST NEIGHBORS

作　　者：郑喜臣 杨易扬 Zheng Xichen;Yang Yiyang(School of Computer Science,Guangdong University of Technology,Guangzhou 510000,Guangdong,China)

机构地区：[1]广东工业大学计算机学院,广东广州510000

出　　处：《计算机应用与软件》2024年第2期264-270,共7页Computer Applications and Software

基　　金：国家自然科学基金项目(61603101)。

摘　　要：针对经典的快速漂移(Quick Shift)算法在偏移过程中需要人为地指定领域值,导致在复杂数据集上表现不佳等问题,提出一种改进的共享近邻密度聚类算法(QS-SNN)。该聚类算法基于共享近邻(SNN),计算出样本点之间的相似度;通过相似度衡量得到样本点的局部密度矩阵;通过在SNN领域中对样本点进行快速偏移,得到最终的聚类结果。在多个数据集上进行实验,结果分析表明,该算法比传统的Quick shift算法以及其他的聚类算法在准确度上有了较大的提升。Aimed at the problem that the Quick Shift algorithm needs to manually specify the field value in the migration process,which leads to poor performance on complex datasets,an improved shared nearest neighbor density clustering algorithm(QS-SNN)is proposed.The proposed algorithm was based on shared nearest neighbors(SNN).It calculated the similarity of each pair of points in the dataset,obtained the local density matrix of the sample points through the similarity measurement.The sample points were quickly shift in the SNN field,so that the final clustering result was obtained.Experiments on multiple data sets show that the QS-SNN algorithm has a greater improvement in accuracy than the traditional Quick shift algorithm and other clustering algorithms.

关 键 词：密度聚类 共享近邻 快速漂移 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术]