基于加权动态时间弯曲的多元时间序列相似性匹配方法
发布时间：2018-10-30 20:14
【摘要】：针对常用方法忽略变量相关性和局部形状特性问题,提出基于加权动态时间弯曲的多元时间序列相似性匹配方法(CPCA-SWDTW).首先,在原加权动态时间弯曲算法基础上,引入形态因子,提出基于形态特征的加权动态时间弯曲算法(SWDTW).然后,提取多元时间序列的主成分作为模式表示,消除变量间的相关性,同时将方差贡献率作为相应主成分的权重.在此基础上,运用SWDTW,度量多元时间序列间的相似度.最后,通过相似性搜索实验表明,CPCA-SWDTW具有较好的准确性和鲁棒性.敏感性分析说明CPCA-SWDTW在一定程度上受到权重函数参数的影响.
[Abstract]:Aiming at the problem of ignoring variable correlation and local shape characteristics in common methods, a multi-variable time series similarity matching method (CPCA-SWDTW) based on weighted dynamic time bending is proposed. First of all, based on the original weighted dynamic time bending algorithm and the introduction of morphological factor, a weighted dynamic time bending algorithm (SWDTW). Based on morphological features is proposed. Then, the principal component of the multivariate time series is extracted as the pattern representation, and the correlation between variables is eliminated, and the contribution rate of variance is taken as the weight of the corresponding principal component. On this basis, SWDTW, is used to measure the similarity between multiple time series. Finally, similarity search experiments show that CPCA-SWDTW has good accuracy and robustness. Sensitivity analysis shows that CPCA-SWDTW is influenced by the parameters of weight function to some extent.
【作者单位】： 国防科学技术大学信息系统与管理学院;
【基金】：国家自然科学基金项目(No.71671186,71571185,71501182)资助~~
【分类号】：TP311.13
[Abstract]:Aiming at the problem of ignoring variable correlation and local shape characteristics in common methods, a multi-variable time series similarity matching method (CPCA-SWDTW) based on weighted dynamic time bending is proposed. First of all, based on the original weighted dynamic time bending algorithm and the introduction of morphological factor, a weighted dynamic time bending algorithm (SWDTW). Based on morphological features is proposed. Then, the principal component of the multivariate time series is extracted as the pattern representation, and the correlation between variables is eliminated, and the contribution rate of variance is taken as the weight of the corresponding principal component. On this basis, SWDTW, is used to measure the similarity between multiple time series. Finally, similarity search experiments show that CPCA-SWDTW has good accuracy and robustness. Sensitivity analysis shows that CPCA-SWDTW is influenced by the parameters of weight function to some extent.
【作者单位】： 国防科学技术大学信息系统与管理学院;
【基金】：国家自然科学基金项目(No.71671186,71571185,71501182)资助~~
【分类号】：TP311.13
【相似文献】
相关期刊论文 前8条
1 徐望明;石汉路;;图像多角度局部特征提取及相似性匹配技术研究[J];电子设计工程;2013年03期
2 饶俊;王太勇;;基于最大团的三维模型相似性匹配方法[J];组合机床与自动化加工技术;2010年10期
3 王红霞;陈俊杰;白炜;王志伟;;二进制粒在旱涝序列相似性匹配中的应用[J];太原理工大学学报;2011年04期
4 赵慧,侯建荣,施伯乐;一种基于分形时变维数的非平稳时间序列相似性匹配方法[J];计算机学报;2005年02期
5 崔晨e，
本文编号：2301056
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2301056.html
