Unsupervised Change Point Detection in Multivariate Time Series

Daoping Wu, Suhas Gundimeda, Shaoshuai Mou, Christopher Quinn
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:3844-3852, 2024.

Abstract

We consider the challenging problem of unsupervised change point detection in multivariate time series when the number of change points is unknown. Our method eliminates the user’s need for careful parameter tuning, enhancing its practicality and usability. Our approach identifies time series segments with similar empirically estimated distributions, coupled with a novel greedy algorithm guided by the minimum description length principle. We provide theoretical guarantees and, through experiments on synthetic and real-world data, provide empirical evidence for its improved performance in identifying meaningful change points in practical settings.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-wu24g, title = {Unsupervised Change Point Detection in Multivariate Time Series}, author = {Wu, Daoping and Gundimeda, Suhas and Mou, Shaoshuai and Quinn, Christopher}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {3844--3852}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/wu24g/wu24g.pdf}, url = {https://proceedings.mlr.press/v238/wu24g.html}, abstract = {We consider the challenging problem of unsupervised change point detection in multivariate time series when the number of change points is unknown. Our method eliminates the user’s need for careful parameter tuning, enhancing its practicality and usability. Our approach identifies time series segments with similar empirically estimated distributions, coupled with a novel greedy algorithm guided by the minimum description length principle. We provide theoretical guarantees and, through experiments on synthetic and real-world data, provide empirical evidence for its improved performance in identifying meaningful change points in practical settings.} }
Endnote
%0 Conference Paper %T Unsupervised Change Point Detection in Multivariate Time Series %A Daoping Wu %A Suhas Gundimeda %A Shaoshuai Mou %A Christopher Quinn %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-wu24g %I PMLR %P 3844--3852 %U https://proceedings.mlr.press/v238/wu24g.html %V 238 %X We consider the challenging problem of unsupervised change point detection in multivariate time series when the number of change points is unknown. Our method eliminates the user’s need for careful parameter tuning, enhancing its practicality and usability. Our approach identifies time series segments with similar empirically estimated distributions, coupled with a novel greedy algorithm guided by the minimum description length principle. We provide theoretical guarantees and, through experiments on synthetic and real-world data, provide empirical evidence for its improved performance in identifying meaningful change points in practical settings.
APA
Wu, D., Gundimeda, S., Mou, S. & Quinn, C.. (2024). Unsupervised Change Point Detection in Multivariate Time Series. Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:3844-3852 Available from https://proceedings.mlr.press/v238/wu24g.html.

Related Material