EWMA Model based Shift-Detection Methods for Detecting Covariate Shifts in Non-Stationary Environments

Haider Raza, G Prasad, Yuhua Li

Research output: Contribution to journalArticle

25 Citations (Scopus)

Abstract

Dataset shift is a very common issue wherein the input data distribution shifts over time in non-stationary environments. A broad range of real-world systems face the challenge of dataset shift. In such systems, continuous monitoring of the process behavior and tracking the state of shift are required in order to decide about initiating adaptive corrections in a timely manner. This paper presents novel methods for covariate shift-detection tests based on a two-stage structure for both univariate and multivariate time-series. The first stage works in an online mode and it uses an exponentially weighted moving average (EWMA) model based control chart to detect the covariate shift-point in non-stationary time-series. The second stage validates the shift-detected by first stage using the Kolmogorov–Smirnov statistical hypothesis test (K–S test) in case of univariate time-series and Hotelling's T-Squared multivariate statistical hypothesis test in case of multivariate time-series. Additionally, several orthogonal transformation and blind source separation algorithms are investigated to counteract the adverse effect of cross-correlation in multivariate time-series on shift-detection performance. The proposed methods are suitable to be run in real-time. Their performance is evaluated through experiments using several synthetic and real-world datasets. Results show that all the covariate shifts are detected with much reduced false-alarms compared to other methods.
LanguageEnglish
Pages659-669
JournalPattern Recognition
Volume48
Issue number3
Early online date5 Aug 2014
DOIs
Publication statusPublished - Mar 2015

Fingerprint

Time series
Statistical tests
Blind source separation
Monitoring
Experiments

Keywords

  • Non-stationary environments
  • Dataset shift-detection
  • Covariate shift
  • EWMA

Cite this

@article{070de6bce6454fe8938fe7e61a3022d0,
title = "EWMA Model based Shift-Detection Methods for Detecting Covariate Shifts in Non-Stationary Environments",
abstract = "Dataset shift is a very common issue wherein the input data distribution shifts over time in non-stationary environments. A broad range of real-world systems face the challenge of dataset shift. In such systems, continuous monitoring of the process behavior and tracking the state of shift are required in order to decide about initiating adaptive corrections in a timely manner. This paper presents novel methods for covariate shift-detection tests based on a two-stage structure for both univariate and multivariate time-series. The first stage works in an online mode and it uses an exponentially weighted moving average (EWMA) model based control chart to detect the covariate shift-point in non-stationary time-series. The second stage validates the shift-detected by first stage using the Kolmogorov–Smirnov statistical hypothesis test (K–S test) in case of univariate time-series and Hotelling's T-Squared multivariate statistical hypothesis test in case of multivariate time-series. Additionally, several orthogonal transformation and blind source separation algorithms are investigated to counteract the adverse effect of cross-correlation in multivariate time-series on shift-detection performance. The proposed methods are suitable to be run in real-time. Their performance is evaluated through experiments using several synthetic and real-world datasets. Results show that all the covariate shifts are detected with much reduced false-alarms compared to other methods.",
keywords = "Non-stationary environments, Dataset shift-detection, Covariate shift, EWMA",
author = "Haider Raza and G Prasad and Yuhua Li",
year = "2015",
month = "3",
doi = "10.1016/j.patcog.2014.07.028",
language = "English",
volume = "48",
pages = "659--669",
journal = "Pattern Recognition",
issn = "0031-3203",
publisher = "Elsevier",
number = "3",

}

EWMA Model based Shift-Detection Methods for Detecting Covariate Shifts in Non-Stationary Environments. / Raza, Haider; Prasad, G; Li, Yuhua.

In: Pattern Recognition, Vol. 48, No. 3, 03.2015, p. 659-669.

Research output: Contribution to journalArticle

TY - JOUR

T1 - EWMA Model based Shift-Detection Methods for Detecting Covariate Shifts in Non-Stationary Environments

AU - Raza, Haider

AU - Prasad, G

AU - Li, Yuhua

PY - 2015/3

Y1 - 2015/3

N2 - Dataset shift is a very common issue wherein the input data distribution shifts over time in non-stationary environments. A broad range of real-world systems face the challenge of dataset shift. In such systems, continuous monitoring of the process behavior and tracking the state of shift are required in order to decide about initiating adaptive corrections in a timely manner. This paper presents novel methods for covariate shift-detection tests based on a two-stage structure for both univariate and multivariate time-series. The first stage works in an online mode and it uses an exponentially weighted moving average (EWMA) model based control chart to detect the covariate shift-point in non-stationary time-series. The second stage validates the shift-detected by first stage using the Kolmogorov–Smirnov statistical hypothesis test (K–S test) in case of univariate time-series and Hotelling's T-Squared multivariate statistical hypothesis test in case of multivariate time-series. Additionally, several orthogonal transformation and blind source separation algorithms are investigated to counteract the adverse effect of cross-correlation in multivariate time-series on shift-detection performance. The proposed methods are suitable to be run in real-time. Their performance is evaluated through experiments using several synthetic and real-world datasets. Results show that all the covariate shifts are detected with much reduced false-alarms compared to other methods.

AB - Dataset shift is a very common issue wherein the input data distribution shifts over time in non-stationary environments. A broad range of real-world systems face the challenge of dataset shift. In such systems, continuous monitoring of the process behavior and tracking the state of shift are required in order to decide about initiating adaptive corrections in a timely manner. This paper presents novel methods for covariate shift-detection tests based on a two-stage structure for both univariate and multivariate time-series. The first stage works in an online mode and it uses an exponentially weighted moving average (EWMA) model based control chart to detect the covariate shift-point in non-stationary time-series. The second stage validates the shift-detected by first stage using the Kolmogorov–Smirnov statistical hypothesis test (K–S test) in case of univariate time-series and Hotelling's T-Squared multivariate statistical hypothesis test in case of multivariate time-series. Additionally, several orthogonal transformation and blind source separation algorithms are investigated to counteract the adverse effect of cross-correlation in multivariate time-series on shift-detection performance. The proposed methods are suitable to be run in real-time. Their performance is evaluated through experiments using several synthetic and real-world datasets. Results show that all the covariate shifts are detected with much reduced false-alarms compared to other methods.

KW - Non-stationary environments

KW - Dataset shift-detection

KW - Covariate shift

KW - EWMA

UR - http://www.sciencedirect.com/science/article/pii/S0031320314002878

U2 - 10.1016/j.patcog.2014.07.028

DO - 10.1016/j.patcog.2014.07.028

M3 - Article

VL - 48

SP - 659

EP - 669

JO - Pattern Recognition

T2 - Pattern Recognition

JF - Pattern Recognition

SN - 0031-3203

IS - 3

ER -