Abstract: Aiming at the problem of multivariate time series similarity search, this paper presents the definition of parameter importance degree and puts forward a candidate sets obtaining method based on it. It extracts singular vector and eigenvector matrix as the features of multivariate time series by SVD, constructs similarity measure modal via coordinate transformation theory in linear space, realizes precise matching on candidate sets and gets ultimate results. Experiments on flight data similarity query show the validity of the method.

Key words: multivariate time series, similarity query, parameter importance degree, feature extraction, similarity measure

摘要： 针对多元时间序列的相似性查询问题，给出参数重要度的定义，提出一种基于参数重要度的候选集查询方法。通过对多元时间序列的SVD分解，将奇异值向量和特征矩阵作为多元序列的特征，基于线性空间中的坐标变换原理构造2个多元时间序列的相似性度量模型，实现在候选集上的精确匹配并获得最终的结果集。对飞行数据的相似性查询实验验证了该方法的有效性。

关键词: 多元时间序列, 相似性查询, 参数重要度, 特征提取, 相似性度量

CLC Number: 

• TP391.4