Predictive handling of asynchronous concept drifts in distributed environments

H.H. Ang, V. Gopalkrishnan, I. Zliobaite, M. Pechenizkiy, S.C.H. Hoi

Research output: Contribution to journalArticleAcademicpeer-review

15 Citations (Scopus)
2 Downloads (Pure)

Abstract

In a distributed computing environment, peers collaboratively learn to classify concepts of interest from each other. When external changes happen and their concepts drift, the peers should adapt to avoid increase in misclassification errors. The problem of adaptation becomes more difficult when the changes are asynchronous, i.e., when peers experience drifts at different times. We address this problem by developing an ensemble approach, PINE, that combines reactive adaptation via drift detection, and proactive handling of upcoming changes via early warning and adaptation across the peers. With empirical study on simulated and real world datasets, we show that PINE handles asynchronous concept drifts better and faster than current state-of-the-art approaches, which have been designed to work in less challenging environments. In addition, PINE is parameter insensitive and incurs less communication cost while achieving better accuracy. Keywords: Classi¿cation, Distributed Systems, Concept Drift.
Original languageEnglish
Pages (from-to)2343-2365
Number of pages13
JournalIEEE Transactions on Knowledge and Data Engineering
Volume25
Issue number10
DOIs
Publication statusPublished - 2013

Fingerprint

Distributed computer systems
Communication
Costs

Cite this

Ang, H.H. ; Gopalkrishnan, V. ; Zliobaite, I. ; Pechenizkiy, M. ; Hoi, S.C.H. / Predictive handling of asynchronous concept drifts in distributed environments. In: IEEE Transactions on Knowledge and Data Engineering. 2013 ; Vol. 25, No. 10. pp. 2343-2365.
@article{a64aecc2acfd4ea68a353ca454d1d992,
title = "Predictive handling of asynchronous concept drifts in distributed environments",
abstract = "In a distributed computing environment, peers collaboratively learn to classify concepts of interest from each other. When external changes happen and their concepts drift, the peers should adapt to avoid increase in misclassification errors. The problem of adaptation becomes more difficult when the changes are asynchronous, i.e., when peers experience drifts at different times. We address this problem by developing an ensemble approach, PINE, that combines reactive adaptation via drift detection, and proactive handling of upcoming changes via early warning and adaptation across the peers. With empirical study on simulated and real world datasets, we show that PINE handles asynchronous concept drifts better and faster than current state-of-the-art approaches, which have been designed to work in less challenging environments. In addition, PINE is parameter insensitive and incurs less communication cost while achieving better accuracy. Keywords: Classi¿cation, Distributed Systems, Concept Drift.",
author = "H.H. Ang and V. Gopalkrishnan and I. Zliobaite and M. Pechenizkiy and S.C.H. Hoi",
year = "2013",
doi = "10.1109/TKDE.2012.172",
language = "English",
volume = "25",
pages = "2343--2365",
journal = "IEEE Transactions on Knowledge and Data Engineering",
issn = "1041-4347",
publisher = "IEEE Computer Society",
number = "10",

}

Predictive handling of asynchronous concept drifts in distributed environments. / Ang, H.H.; Gopalkrishnan, V.; Zliobaite, I.; Pechenizkiy, M.; Hoi, S.C.H.

In: IEEE Transactions on Knowledge and Data Engineering, Vol. 25, No. 10, 2013, p. 2343-2365.

Research output: Contribution to journalArticleAcademicpeer-review

TY - JOUR

T1 - Predictive handling of asynchronous concept drifts in distributed environments

AU - Ang, H.H.

AU - Gopalkrishnan, V.

AU - Zliobaite, I.

AU - Pechenizkiy, M.

AU - Hoi, S.C.H.

PY - 2013

Y1 - 2013

N2 - In a distributed computing environment, peers collaboratively learn to classify concepts of interest from each other. When external changes happen and their concepts drift, the peers should adapt to avoid increase in misclassification errors. The problem of adaptation becomes more difficult when the changes are asynchronous, i.e., when peers experience drifts at different times. We address this problem by developing an ensemble approach, PINE, that combines reactive adaptation via drift detection, and proactive handling of upcoming changes via early warning and adaptation across the peers. With empirical study on simulated and real world datasets, we show that PINE handles asynchronous concept drifts better and faster than current state-of-the-art approaches, which have been designed to work in less challenging environments. In addition, PINE is parameter insensitive and incurs less communication cost while achieving better accuracy. Keywords: Classi¿cation, Distributed Systems, Concept Drift.

AB - In a distributed computing environment, peers collaboratively learn to classify concepts of interest from each other. When external changes happen and their concepts drift, the peers should adapt to avoid increase in misclassification errors. The problem of adaptation becomes more difficult when the changes are asynchronous, i.e., when peers experience drifts at different times. We address this problem by developing an ensemble approach, PINE, that combines reactive adaptation via drift detection, and proactive handling of upcoming changes via early warning and adaptation across the peers. With empirical study on simulated and real world datasets, we show that PINE handles asynchronous concept drifts better and faster than current state-of-the-art approaches, which have been designed to work in less challenging environments. In addition, PINE is parameter insensitive and incurs less communication cost while achieving better accuracy. Keywords: Classi¿cation, Distributed Systems, Concept Drift.

U2 - 10.1109/TKDE.2012.172

DO - 10.1109/TKDE.2012.172

M3 - Article

VL - 25

SP - 2343

EP - 2365

JO - IEEE Transactions on Knowledge and Data Engineering

JF - IEEE Transactions on Knowledge and Data Engineering

SN - 1041-4347

IS - 10

ER -