基于希尔伯特相似度的高维面板数据聚类方法及应用  

Hilbert Similarity-Based High-dimensional Panel Data Clustering Method and Its Application

作　　者：袁欣 俞卫琴 王国强 Yuan Xin;Yu Weiqin;Wang Guoqiang(School of Mathematics,Physics and Statistics,Shanghai University of Engineering Science,Shanghai 201620,China)

机构地区：[1]上海工程技术大学数理与统计学院,上海201620

出　　处：《统计与决策》2022年第17期52-54,共3页Statistics & Decision

基　　金：国家自然科学基金资助项目(11971302)。

摘　　要：针对面板数据的高维度性,文章将希尔伯特相似度引入聚类分析方法中,将高维面板数据映射到希尔伯特空间上并构造希尔伯特指数,计算该指数的离散概率分布并进行可视化展示,并基于Jensen-Shannon距离进行系统聚类。实证结果表明,基于希尔伯特相似度的聚类方法更适用于高维面板数据,聚类结果稳定且可视化效果较好。In view of the high dimensionality of panel data,this paper brings Hilbert similarity into clustering analysis method.Firstly,the high-dimensional panel data is mapped onto the Hilbert space and the Hilbert index is constructed.Secondly,the discrete probability distribution of the index is calculated and visualized.Finally,the system is clustered based on Jensen-Shannon distance.The empirical results show that the clustering method based on Hilbert similarity is more suitable for high-dimensional panel data,and that the clustering results are stable and the visualization is better.

关 键 词：高维面板数据 希尔伯特相似度 Jensen-Shannon距离 聚类分析 

分 类 号：C812[社会学—统计学]