Home > Research > Publications & Outputs > Collective Anomaly Detection in High-Dimensiona...

Electronic data

  • SS-2021-0181_na

    Accepted author manuscript, 3.57 MB, PDF document

Links

Text available via DOI:

View graph of relations

Collective Anomaly Detection in High-Dimensional Var Models

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Published

Standard

Collective Anomaly Detection in High-Dimensional Var Models. / Maeng, Hyeyoung; Eckley, Idris; Fearnhead, Paul.
In: Statistica Sinica, Vol. 33, No. 1603-1627, 31.05.2023, p. 1603-1627.

Research output: Contribution to Journal/MagazineJournal articlepeer-review

Harvard

APA

Vancouver

Maeng H, Eckley I, Fearnhead P. Collective Anomaly Detection in High-Dimensional Var Models. Statistica Sinica. 2023 May 31;33(1603-1627):1603-1627. doi: 10.5705/ss.202021.0181

Author

Maeng, Hyeyoung ; Eckley, Idris ; Fearnhead, Paul. / Collective Anomaly Detection in High-Dimensional Var Models. In: Statistica Sinica. 2023 ; Vol. 33, No. 1603-1627. pp. 1603-1627.

Bibtex

@article{15f25fc1adf14a5694c20ae608c9945a,
title = "Collective Anomaly Detection in High-Dimensional Var Models",
abstract = "There is increasing interest in detecting collective anomalies: potentially short periods during which the features of data change, before reverting back to normal behavior. We propose a new method for detecting a collective anomaly in the vector autoregressive (VAR) models. We focus on situations in which the change in the VAR coefficient matrix at an anomaly is sparse, that is, a small number of entries of the VAR coefficient matrix change. To tackle this problem, we propose a test statistic for a local segment that is built on the lasso estimator of the change in the model parameters. This enables us to detect a sparse change more efficiently, and our lasso-based approach becomes especially advantageous when the anomalous interval is short. We show that the new procedure controls the type-I error and has asymptotic power tending to one. The practicality of our approach is demonstrated using simulations and two data examples, involving New York taxi trip data and EEG data, respectively.",
author = "Hyeyoung Maeng and Idris Eckley and Paul Fearnhead",
year = "2023",
month = may,
day = "31",
doi = "10.5705/ss.202021.0181",
language = "English",
volume = "33",
pages = "1603--1627",
journal = "Statistica Sinica",
issn = "1017-0405",
publisher = "Institute of Statistical Science",
number = "1603-1627",

}

RIS

TY - JOUR

T1 - Collective Anomaly Detection in High-Dimensional Var Models

AU - Maeng, Hyeyoung

AU - Eckley, Idris

AU - Fearnhead, Paul

PY - 2023/5/31

Y1 - 2023/5/31

N2 - There is increasing interest in detecting collective anomalies: potentially short periods during which the features of data change, before reverting back to normal behavior. We propose a new method for detecting a collective anomaly in the vector autoregressive (VAR) models. We focus on situations in which the change in the VAR coefficient matrix at an anomaly is sparse, that is, a small number of entries of the VAR coefficient matrix change. To tackle this problem, we propose a test statistic for a local segment that is built on the lasso estimator of the change in the model parameters. This enables us to detect a sparse change more efficiently, and our lasso-based approach becomes especially advantageous when the anomalous interval is short. We show that the new procedure controls the type-I error and has asymptotic power tending to one. The practicality of our approach is demonstrated using simulations and two data examples, involving New York taxi trip data and EEG data, respectively.

AB - There is increasing interest in detecting collective anomalies: potentially short periods during which the features of data change, before reverting back to normal behavior. We propose a new method for detecting a collective anomaly in the vector autoregressive (VAR) models. We focus on situations in which the change in the VAR coefficient matrix at an anomaly is sparse, that is, a small number of entries of the VAR coefficient matrix change. To tackle this problem, we propose a test statistic for a local segment that is built on the lasso estimator of the change in the model parameters. This enables us to detect a sparse change more efficiently, and our lasso-based approach becomes especially advantageous when the anomalous interval is short. We show that the new procedure controls the type-I error and has asymptotic power tending to one. The practicality of our approach is demonstrated using simulations and two data examples, involving New York taxi trip data and EEG data, respectively.

U2 - 10.5705/ss.202021.0181

DO - 10.5705/ss.202021.0181

M3 - Journal article

VL - 33

SP - 1603

EP - 1627

JO - Statistica Sinica

JF - Statistica Sinica

SN - 1017-0405

IS - 1603-1627

ER -