A fast sorting-based aggregation method for symbolic time series representation
Xinye Chen, Stefan Güttel · 2021 International Conference on Data Mining Workshops (ICDMW) · 2021
A new variant of the adaptive Brownian bridge-based approximation (ABBA) for the symbolic representation of time series is presented. This variant, called fABBA, utilizes a new aggregation approach tailored to the piecewise representation of time series. By replacing the k-means clustering used in ABBA with a sorting-based aggregation technique, and thereby avoiding repeated within-cluster-sum-of-squares computations, the computational complexity is significantly reduced. Also, in contrast to the original variant, the new approach does not require the number of time series symbols to be specified in advance. Through extensive tests and using performance profiles, we demonstrate that the new method significantly outperforms ABBA. In terms of reconstruction quality, we find that fABBA remains competitive to ABBA, despite the significant reduction in computation time, while it outperforms the popular SAX and 1d-SAX representations of time series.