文献类型：期刊文献

中文题名：基于改进密度峰值算法的轨迹聚类

英文题名：Trajectory clustering based on improved density peak algorithm

作者：钟超[1];刘漫丹[1];贺帆[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2024

卷号：45

期号：1

起止页码：130

中文期刊名：计算机工程与设计

外文期刊名：Computer Engineering and Design

收录：CSTPCD;;北大核心:【北大核心2023】；

语种：中文

中文关键词：无线网络;密度峰值聚类;语义信息;相似性度量;信息熵;聚类中心;共享近邻贡献度

外文关键词：wireless network;density peak clustering;semantic information;similarity measure;information entropy;clustering centers;shared nearest neighbor contribution

摘要：为解决用户群体移动轨迹划分和密度峰值聚类算法自身局限性的问题,以校园轨迹为对象,考虑时间和位置语义信息层面的信息,建立网络用户间的相似性度量模型,提出一种基于共享近邻贡献度的密度峰值聚类算法(density peak clustering based on shared nearest neighbor contribution,SNNC-DPC),结合信息熵理论,通过最小化局部密度熵自适应选择截断距离;在局部密度计算上,利用共享近邻贡献度重新计算局部密度,更加全面地反映数据分布的特性;采用非线性变换方法选取决策值,解决聚类中心选取困难且方法单一的问题。在真实校园轨迹数据集上实验,验证了改进算法的有效性。
To solve the problems of the limitations of user group moving trajectory division and density peak clustering algorithm,taking the campus trajectory as the object,considering the information of time and location semantic information,a similarity measurement model among network users was established,and a density peak clustering based on shared nearest neighbor contribution(SNNC-DPC)algorithm was proposed,in which the information entropy theory was combined,the truncation distance was adaptively selected by minimizing the local density entropy.In the local density calculation,the shared nearest neighbor contribution was used to recalculate the local density,which reflected the characteristics of data distribution more comprehensively.The decision value was selected using the nonlinear transformation method to solve the problem of difficult and single method for selecting the clustering center.Experiments on real campus trajectory data sets verify the effectiveness of the improved algorithm.