Smart sampling: a novel unsupervised boosting approach for outlier detection

Mahsa Salehi, Xuyun Zhang, James C. Bezdek, Christopher Leckie

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearch

Abstract

While various ensemble algorithms have been proposed for supervised ensembles or clustering ensembles, there are few ensemble based approaches for outlier detection. The main challenge in this context is the lack of knowledge about the accuracy of the outlier detectors. Hence, none of the proposed approaches focused on sequential boosting techniques. In this paper for the first time we propose a novel boosting algorithm for outlier detection called BSS, where we sequentially improve the accuracy of each ensemble detector in an unsupervised manner. We discuss the effectiveness of our approach in terms of bias-variance tradeoff. Furthermore, an extended version of BSS (called DBSS) is proposed to introduce a novel source of diversity in outlier ensemble modeling. DBSS is used to analyze the effect of changing the input parameter of BSS on its detection accuracy. Our experimental results on both synthetic and real data sets demonstrate that our approaches outperform the two state-of-the-art outlier ensemble algorithms and benefit from bias reduction. In addition, our BSS approach is robust with respect to the changing input parameter. Since each detector in our proposed BSS/DBSS is only a subset of the whole dataset, our both techniques are well suited to application environments with limited memory processors (e.g., wireless sensor networks).

Original languageEnglish
Title of host publicationAI 2016: Advances in Artificial Intelligence
Subtitle of host publication29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings
EditorsByeong Ho Kang, Quan Bai
Place of PublicationCham Switzerland
PublisherSpringer
Pages469-481
Number of pages13
ISBN (Electronic)9783319501277
ISBN (Print)9783319501260
DOIs
Publication statusPublished - 2016
Externally publishedYes
EventAustralasian Joint Conference on Artificial Intelligence 2016 - Hobart, Australia
Duration: 5 Dec 20168 Dec 2016
Conference number: 29th
https://ai2016.net/

Publication series

NameLecture Notes in Artificial Intelligence
PublisherSpringer
Volume9992
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceAustralasian Joint Conference on Artificial Intelligence 2016
Abbreviated titleAI 2016
CountryAustralia
CityHobart
Period5/12/168/12/16
Internet address

Keywords

  • Ensemble analysis
  • Outlier detection
  • Smart sampling
  • Unsupervised boosting

Cite this

Salehi, M., Zhang, X., Bezdek, J. C., & Leckie, C. (2016). Smart sampling: a novel unsupervised boosting approach for outlier detection. In B. H. Kang, & Q. Bai (Eds.), AI 2016: Advances in Artificial Intelligence: 29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings (pp. 469-481). (Lecture Notes in Artificial Intelligence; Vol. 9992). Cham Switzerland: Springer. https://doi.org/10.1007/978-3-319-50127-7_40
Salehi, Mahsa ; Zhang, Xuyun ; Bezdek, James C. ; Leckie, Christopher. / Smart sampling : a novel unsupervised boosting approach for outlier detection. AI 2016: Advances in Artificial Intelligence: 29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings. editor / Byeong Ho Kang ; Quan Bai. Cham Switzerland : Springer, 2016. pp. 469-481 (Lecture Notes in Artificial Intelligence).
@inproceedings{5984870d784f48bbb1b241079daa1f11,
title = "Smart sampling: a novel unsupervised boosting approach for outlier detection",
abstract = "While various ensemble algorithms have been proposed for supervised ensembles or clustering ensembles, there are few ensemble based approaches for outlier detection. The main challenge in this context is the lack of knowledge about the accuracy of the outlier detectors. Hence, none of the proposed approaches focused on sequential boosting techniques. In this paper for the first time we propose a novel boosting algorithm for outlier detection called BSS, where we sequentially improve the accuracy of each ensemble detector in an unsupervised manner. We discuss the effectiveness of our approach in terms of bias-variance tradeoff. Furthermore, an extended version of BSS (called DBSS) is proposed to introduce a novel source of diversity in outlier ensemble modeling. DBSS is used to analyze the effect of changing the input parameter of BSS on its detection accuracy. Our experimental results on both synthetic and real data sets demonstrate that our approaches outperform the two state-of-the-art outlier ensemble algorithms and benefit from bias reduction. In addition, our BSS approach is robust with respect to the changing input parameter. Since each detector in our proposed BSS/DBSS is only a subset of the whole dataset, our both techniques are well suited to application environments with limited memory processors (e.g., wireless sensor networks).",
keywords = "Ensemble analysis, Outlier detection, Smart sampling, Unsupervised boosting",
author = "Mahsa Salehi and Xuyun Zhang and Bezdek, {James C.} and Christopher Leckie",
year = "2016",
doi = "10.1007/978-3-319-50127-7_40",
language = "English",
isbn = "9783319501260",
series = "Lecture Notes in Artificial Intelligence",
publisher = "Springer",
pages = "469--481",
editor = "Kang, {Byeong Ho } and Quan Bai",
booktitle = "AI 2016: Advances in Artificial Intelligence",

}

Salehi, M, Zhang, X, Bezdek, JC & Leckie, C 2016, Smart sampling: a novel unsupervised boosting approach for outlier detection. in BH Kang & Q Bai (eds), AI 2016: Advances in Artificial Intelligence: 29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings. Lecture Notes in Artificial Intelligence, vol. 9992, Springer, Cham Switzerland, pp. 469-481, Australasian Joint Conference on Artificial Intelligence 2016, Hobart, Australia, 5/12/16. https://doi.org/10.1007/978-3-319-50127-7_40

Smart sampling : a novel unsupervised boosting approach for outlier detection. / Salehi, Mahsa; Zhang, Xuyun; Bezdek, James C.; Leckie, Christopher.

AI 2016: Advances in Artificial Intelligence: 29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings. ed. / Byeong Ho Kang; Quan Bai. Cham Switzerland : Springer, 2016. p. 469-481 (Lecture Notes in Artificial Intelligence; Vol. 9992).

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearch

TY - GEN

T1 - Smart sampling

T2 - a novel unsupervised boosting approach for outlier detection

AU - Salehi, Mahsa

AU - Zhang, Xuyun

AU - Bezdek, James C.

AU - Leckie, Christopher

PY - 2016

Y1 - 2016

N2 - While various ensemble algorithms have been proposed for supervised ensembles or clustering ensembles, there are few ensemble based approaches for outlier detection. The main challenge in this context is the lack of knowledge about the accuracy of the outlier detectors. Hence, none of the proposed approaches focused on sequential boosting techniques. In this paper for the first time we propose a novel boosting algorithm for outlier detection called BSS, where we sequentially improve the accuracy of each ensemble detector in an unsupervised manner. We discuss the effectiveness of our approach in terms of bias-variance tradeoff. Furthermore, an extended version of BSS (called DBSS) is proposed to introduce a novel source of diversity in outlier ensemble modeling. DBSS is used to analyze the effect of changing the input parameter of BSS on its detection accuracy. Our experimental results on both synthetic and real data sets demonstrate that our approaches outperform the two state-of-the-art outlier ensemble algorithms and benefit from bias reduction. In addition, our BSS approach is robust with respect to the changing input parameter. Since each detector in our proposed BSS/DBSS is only a subset of the whole dataset, our both techniques are well suited to application environments with limited memory processors (e.g., wireless sensor networks).

AB - While various ensemble algorithms have been proposed for supervised ensembles or clustering ensembles, there are few ensemble based approaches for outlier detection. The main challenge in this context is the lack of knowledge about the accuracy of the outlier detectors. Hence, none of the proposed approaches focused on sequential boosting techniques. In this paper for the first time we propose a novel boosting algorithm for outlier detection called BSS, where we sequentially improve the accuracy of each ensemble detector in an unsupervised manner. We discuss the effectiveness of our approach in terms of bias-variance tradeoff. Furthermore, an extended version of BSS (called DBSS) is proposed to introduce a novel source of diversity in outlier ensemble modeling. DBSS is used to analyze the effect of changing the input parameter of BSS on its detection accuracy. Our experimental results on both synthetic and real data sets demonstrate that our approaches outperform the two state-of-the-art outlier ensemble algorithms and benefit from bias reduction. In addition, our BSS approach is robust with respect to the changing input parameter. Since each detector in our proposed BSS/DBSS is only a subset of the whole dataset, our both techniques are well suited to application environments with limited memory processors (e.g., wireless sensor networks).

KW - Ensemble analysis

KW - Outlier detection

KW - Smart sampling

KW - Unsupervised boosting

UR - http://www.scopus.com/inward/record.url?scp=85007211652&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-50127-7_40

DO - 10.1007/978-3-319-50127-7_40

M3 - Conference Paper

SN - 9783319501260

T3 - Lecture Notes in Artificial Intelligence

SP - 469

EP - 481

BT - AI 2016: Advances in Artificial Intelligence

A2 - Kang, Byeong Ho

A2 - Bai, Quan

PB - Springer

CY - Cham Switzerland

ER -

Salehi M, Zhang X, Bezdek JC, Leckie C. Smart sampling: a novel unsupervised boosting approach for outlier detection. In Kang BH, Bai Q, editors, AI 2016: Advances in Artificial Intelligence: 29th Australasian Joint Conference, Hobart, TAS, Australia, December 5-8, 2016, Proceedings. Cham Switzerland: Springer. 2016. p. 469-481. (Lecture Notes in Artificial Intelligence). https://doi.org/10.1007/978-3-319-50127-7_40