The MTSR package was used to verify the effectiveness of the solution proposed in 1, in terms of processing speed and RAM consumption (on generated dataset), and in terms of subsequent cluster validity (on real UK smart meter data 2). Please note that this implementation is made as a proof of concept, and not as a complete production-ready package. However, if you are using this software, please cite reference 1.
Prerequisites:
You can also try an extended solution from the culearn package, written entirely in Python.
Footnotes
-
Igor Manojlović, Goran Švenda, Aleksandar Erdeljan, Milan Gavrić, Darko Čapko: Hierarchical multiresolution representation of streaming time series, Big Data Research 26: 100256 (2021), DOI: 10.1016/j.bdr.2021.100256 ↩ ↩2
-
REFIT dataset, https://pureportal.strath.ac.uk/en/datasets/refit-electrical-load-measurements-cleaned ↩