Classification of Multivariate Time Series Using Supervised Locality Preserving Projection

Abstract

Multivariate time series (MTS) are used in very broad areas such as finance, medicine, multimedia and speech recognition. Most of existing approaches for MTS classification are not designed for preserving the within-class local structure of the MTS dataset. The within-class local structure is important when a classifier is used for classification. In this paper, a new feature extraction method for MTS classification based on supervised Locality Preserving Projection (LPP) is proposed. MTS samples are projected into the PCA (principal component analysis) subspace by throwing away the smallest principal components, and then, the MTS samples in the PCA subspace are projected into a lower-dimensional space by using supervised LPP. Experimental results performed on five real-world datasets demonstrate the effectiveness of our proposed approach for MTS classification.