A novel trend based SAX reduction technique for time series

摘要

We propose in this paper a novel trend based SAX reduction technique called SAX_CP, which captures the trends in a time series based on abrupt change points and on data variation. SAX_CP discriminates between time series by adopting a variable segment size that depends on the time series' change points. Furthermore, SAX_CP is endowed with a new distance that takes into account the possible correlations between time series. The proposed distance enjoys interesting features such as lower bounding the Euclidean distance and better Tightness of Lower Bound (TLB) compared to the so far published SAX based distances. We conduct comprehensive experiments on classification based on SAX_CP and other related SAX based reduction techniques. The results show that SAX_CP enjoys better classification results compared to these techniques with a small reduction cost. (C) 2019 Elsevier Ltd. All rights reserved.