Multivariate Time Series Data Mining: PCA-based Measures For SimilaritySearch

Abstract

In this paper, we discuss the application of Principal Component Analysis (PCA), for the purpose of determining a similarity/distance measure among multivariate time series. We review several PCA-based measures that have been proposed by researchers from diverse scientific fields and we extend the well-known statistic in the Statistical Process Control community, SPE, in order to define a novel distance measure. We conducted experiments on four datasets, which have been used extensively in the literature, and we provide the results of their performance with respect to classification accuracy. Experiments indicate that there is no measure that can be clearly considered as the most appropriate one for any dataset, and that the newly proposed measure is a promising option for similarity search.