基于加权范数的多维时间序列相似性主元分析

Similarity PCA of multivariate time series based on extended Frobenius norm

为提高多维时间序列相似性搜索的效率,利用多维时间序列的协方差矩阵的特征值和特征向量构造加权Frobe-nius范数,将其作为多维时间序列主元之间距离,并将其用于对多维时间序列主元相似度的度量.在相似性搜索算法中分别采用不同的相似性度量方法作比较.实验结果表明,相对于其他的传统多维时间序列相似性度量方法,这种基于加权Frobenius范数的方法在查全率和查准率上具有更大的优越性.

To improve the efficiency of similarity search of Multivariate Time Series（MTS）,Extended Frobenius norm（Eros） consisting of eigen value and eigen vector of MTS covariance matrix is adopted for similarity Principal Component Analysis（PCA） of MTS datasets.Base on different similarity measurements methods,several similarity searching experiments are carried out to show the validity of the method.The results indicate that Eros has superiority in recall precision as compared to the traditional similarity measures for MTS datasets.