Home > Research > Publications & Outputs > Automatic Change-Point Detection in Time Series...

Electronic data

  • 2211.03860v1

    Accepted author manuscript, 1.06 MB, PDF document

    Available under license: CC BY: Creative Commons Attribution 4.0 International License

Text available via DOI:

View graph of relations

Automatic Change-Point Detection in Time Series via Deep Learning

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Published

Standard

Automatic Change-Point Detection in Time Series via Deep Learning. / Li, Jie; Fearnhead, Paul; Fryzlewicz, Piotr et al.
In: Journal of the Royal Statistical Society: Series B (Statistical Methodology), Vol. 86, No. 2, 12.04.2024, p. 273-285.

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Harvard

Li, J, Fearnhead, P, Fryzlewicz, P & Wang, T 2024, 'Automatic Change-Point Detection in Time Series via Deep Learning', Journal of the Royal Statistical Society: Series B (Statistical Methodology), vol. 86, no. 2, pp. 273-285. https://doi.org/10.1093/jrsssb/qkae004

APA

Li, J., Fearnhead, P., Fryzlewicz, P., & Wang, T. (2024). Automatic Change-Point Detection in Time Series via Deep Learning. Journal of the Royal Statistical Society: Series B (Statistical Methodology), 86(2), 273-285. https://doi.org/10.1093/jrsssb/qkae004

Vancouver

Li J, Fearnhead P, Fryzlewicz P, Wang T. Automatic Change-Point Detection in Time Series via Deep Learning. Journal of the Royal Statistical Society: Series B (Statistical Methodology). 2024 Apr 12;86(2):273-285. Epub 2024 Jan 11. doi: 10.1093/jrsssb/qkae004

Author

Li, Jie ; Fearnhead, Paul ; Fryzlewicz, Piotr et al. / Automatic Change-Point Detection in Time Series via Deep Learning. In: Journal of the Royal Statistical Society: Series B (Statistical Methodology). 2024 ; Vol. 86, No. 2. pp. 273-285.

Bibtex

@article{53bf8ed04dd545c4a2c5447179ac1185,
title = "Automatic Change-Point Detection in Time Series via Deep Learning",
abstract = "Detecting change points in data is challenging because of the range of possible types of change and types of behaviour of data when there is no change. Statistically efficient methods for detecting a change will depend on both of these features, and it can be difficult for a practitioner to develop an appropriate detection method for their application of interest. We show how to automatically generate new offline detection methods based on training a neural network. Our approach is motivated by many existing tests for the presence of a change point being representable by a simple neural network, and thus a neural network trained with sufficient data should have performance at least as good as these methods. We present theory that quantifies the error rate for such an approach, and how it depends on the amount of training data. Empirical results show that, even with limited training data, its performance is competitive with the standard cumulative sum (CUSUM) based classifier for detecting a change in mean when the noise is independent and Gaussian, and can substantially outperform it in the presence of auto-correlated or heavy-tailed noise. Our method also shows strong results in detecting and localizing changes in activity based on accelerometer data.",
keywords = "Statistics and Probability, Statistics, Probability and Uncertainty",
author = "Jie Li and Paul Fearnhead and Piotr Fryzlewicz and Tengyao Wang",
year = "2024",
month = apr,
day = "12",
doi = "10.1093/jrsssb/qkae004",
language = "English",
volume = "86",
pages = "273--285",
journal = "Journal of the Royal Statistical Society: Series B (Statistical Methodology)",
issn = "1369-7412",
publisher = "Wiley-Blackwell",
number = "2",

}

RIS

TY - JOUR

T1 - Automatic Change-Point Detection in Time Series via Deep Learning

AU - Li, Jie

AU - Fearnhead, Paul

AU - Fryzlewicz, Piotr

AU - Wang, Tengyao

PY - 2024/4/12

Y1 - 2024/4/12

N2 - Detecting change points in data is challenging because of the range of possible types of change and types of behaviour of data when there is no change. Statistically efficient methods for detecting a change will depend on both of these features, and it can be difficult for a practitioner to develop an appropriate detection method for their application of interest. We show how to automatically generate new offline detection methods based on training a neural network. Our approach is motivated by many existing tests for the presence of a change point being representable by a simple neural network, and thus a neural network trained with sufficient data should have performance at least as good as these methods. We present theory that quantifies the error rate for such an approach, and how it depends on the amount of training data. Empirical results show that, even with limited training data, its performance is competitive with the standard cumulative sum (CUSUM) based classifier for detecting a change in mean when the noise is independent and Gaussian, and can substantially outperform it in the presence of auto-correlated or heavy-tailed noise. Our method also shows strong results in detecting and localizing changes in activity based on accelerometer data.

AB - Detecting change points in data is challenging because of the range of possible types of change and types of behaviour of data when there is no change. Statistically efficient methods for detecting a change will depend on both of these features, and it can be difficult for a practitioner to develop an appropriate detection method for their application of interest. We show how to automatically generate new offline detection methods based on training a neural network. Our approach is motivated by many existing tests for the presence of a change point being representable by a simple neural network, and thus a neural network trained with sufficient data should have performance at least as good as these methods. We present theory that quantifies the error rate for such an approach, and how it depends on the amount of training data. Empirical results show that, even with limited training data, its performance is competitive with the standard cumulative sum (CUSUM) based classifier for detecting a change in mean when the noise is independent and Gaussian, and can substantially outperform it in the presence of auto-correlated or heavy-tailed noise. Our method also shows strong results in detecting and localizing changes in activity based on accelerometer data.

KW - Statistics and Probability

KW - Statistics, Probability and Uncertainty

U2 - 10.1093/jrsssb/qkae004

DO - 10.1093/jrsssb/qkae004

M3 - Journal article

VL - 86

SP - 273

EP - 285

JO - Journal of the Royal Statistical Society: Series B (Statistical Methodology)

JF - Journal of the Royal Statistical Society: Series B (Statistical Methodology)

SN - 1369-7412

IS - 2

ER -