统计与决策2025，Vol.41Issue(22)：24-30,7.DOI:10.13546/j.cnki.tjyjc.2025.22.004
基于GVB-CEEMD-DWT的数据变点检测方法及应用
Data Change Point Detection Method Based on GVB-CEEMD-DWT and Its Application
摘要
Abstract
Change point detection in data is of significant value for analyzing sequence change patterns and studying future evolution trends.This paper combines data processing methods such as Gaussian Blur,Complete Ensemble Empirical Mode De-composition,and Wavelet Transform,leverages the advantages of different approaches to construct a multi-modal combined detec-tion method for studying change-point detection in data sequences.An empirical study is conducted based on the data of 243 trad-ing days of China's Shanghai Composite Index from January 4,2022 to December 30,2022,and the results are shown as follow:As for the principal factor sequence after Gaussian blur and modal decomposition,due to the improvement of the signal-to-noise ratio,the real change patterns in the data are more clearly displayed,enabling the subsequent wavelet transform to more accurately extract the key features in the data generation process and improving the accuracy of change point detection.High-frequency de-tail coefficients at different resolution levels exhibit significant differences in sensitivity to sequence changes.Detail coefficients at higher resolution levels can reveal more subtle changes in the sequence and identify change points,while those at lower resolution levels are unable to reflect such fine variations and lacking in the ability to detect change points.The fluctuations in stock market return sequences primarily stem from random market disturbances,with the long-term trend component having an almost negligi-ble impact on sequence volatility.To some extent,stock prices are not predictable.关键词
正态卷积核/高斯模糊/完全聚合经验模态分解/小波变换/变点检测Key words
normal convolution kernel/Gaussian blur/complete ensemble empirical mode decomposition/wavelet transform/change-point detection分类
社会科学引用本文复制引用
郭建平..基于GVB-CEEMD-DWT的数据变点检测方法及应用[J].统计与决策,2025,41(22):24-30,7.基金项目
国家社会科学基金一般项目(22BTJ061) （22BTJ061）
