TIME-SERIES SEGMENTATION - A SLIDING WINDOW APPROACH
TIME-SERIES SEGMENTATION - A SLIDING WINDOW APPROACH
复制标题
DOI：
10.1016/0020-0255(95)00021-g
复制
发表时间：
1995-07-01
影响因子：
8.1
通讯作者：
CHU, CSJ
中科院分区：
文献类型：
--
作者：
CHU, CSJ
本文的目的是提出两个在线，滑动窗口分割算法。检测非平稳性是基于参数波动和赤池信息准则的变点定位。分析了算法的渐近性质。具体地说，极限分布推导和渐近阈值制成表格，以供将来参考。有限样本的模拟来说明这些算法的实用性。
The aim of this paper is to present two on-line, sliding window segmentation algorithms. Detection nonstationarity is based on parameter fluctuations and change point localization of the Akaike information criterion. Asymptotic properties of the proposed algorithms are analyzed. Specifically, the limiting distributions are derived and the asymptotic threshold values are tabulated for future reference. Finite sample simulations are performed to illustrate the usefulness of these algorithms.