桂林理工大学学报2025，Vol.45Issue(2)：279-284,6.DOI:10.3969/j.issn.1674-9057.2025.02.017
基于改进的加权动态时间规整的面板数据聚类方法
Panel data clustering method based on improved weighted dynamic time warping
摘要
Abstract
When dynamic time warping(DTW)is introduced to measure the similarity between samples in the process of panel data clustering,unreasonable matching is easy to occur and the shape similarity of data is not considered in the matching process.To address this issue,this article adopts a method that combines Euclidean distance and slope to construct a distance function for weighted dynamic time warping(WDTW),and utilizes existing clustering algorithms to cluster panel data.This method can effectively improve the low accuracy of clustering caused by excessive stretching or compression in sequence matching,and comprehensively consider the numerical similarity and shape similarity of data in the matching process.To obtain the optimal clustering re-sults,it adjusts the weights of different numerical and shape values to adapt to different datasets.Numerical simulation results show that this method can effectively improve the accuracy of clustering.It can be seen from the clustering results of the empirical analysis that the method can achieve reasonable clustering of panel data and make it more suitable for the actual situation.关键词
面板数据/主成分分析/加权动态时间规整/形状相似性/聚类评价指标Key words
panel data/principal component analysis/weighted dynamic time warping/shape similarity/clus-tering evaluation index分类
数理科学引用本文复制引用
韩柳沅,邓光明..基于改进的加权动态时间规整的面板数据聚类方法[J].桂林理工大学学报,2025,45(2):279-284,6.基金项目
国家自然科学基金项目(71963008) （71963008）
