Early Classification of Multivariate Time Series Based on Piecewise Aggregate Approximation

Chaohong Ma, Xiaoqing Weng, Zhongnan Shan

Abstract: Early Classification on Time Series is becoming more significant in the field of Time Series Data Ming. Especially in some time-sensitive filed, it is obviously preferred to make earlier classification, such as Medical science, Health informatics et al. However, the research tasks are mainly focused on UTS, those of MTS are less. MTS is faced with variable-based and time-based dimensionality. It is significant to find appropriate dimensionality reduction in the practical application of early classification on multivariate time series. We propose a novel method MTEECP based on center sequence and Piecewise Aggregate Approximation which achieve early classification in low-dimension space. Experimental results on 6 real datasets intuitively show our proposed method can reach favorable early classification on MTS.