Multiple distribution data description learning algorithm for novelty detection

Trung Le; Dat Tran; Wanli Ma; Dharmendra Sharma

doi:10.1007/978-3-642-20847-8-21

Multiple distribution data description learning algorithm for novelty detection

Trung Le, Dat Tran, Wanli Ma, Dharmendra Sharma

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research › peer-review

5 Citations (Scopus)

Abstract

Current data description learning methods for novelty detection such as support vector data description and small sphere with large margin construct a spherically shaped boundary around a normal data set to separate this set from abnormal data. The volume of this sphere is minimized to reduce the chance of accepting abnormal data. However those learning methods do not guarantee that the single spherically shaped boundary can best describe the normal data set if there exist some distinctive data distributions in this set. We propose in this paper a new data description learning method that constructs a set of spherically shaped boundaries to provide a better data description to the normal data set. An optimisation problem is proposed and solving this problem results in an iterative learning algorithm to determine the set of spherically shaped boundaries. We prove that the classification error will be reduced after each iteration in our learning method. Experimental results on 28 well-known data sets show that the proposed method provides lower classification error rates.

Original language	English
Title of host publication	Advances in Knowledge Discovery and Data Mining
Subtitle of host publication	15th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 201, Part II
Publisher	Springer
Pages	246-257
Number of pages	12
ISBN (Print)	9783642208461
DOIs	https://doi.org/10.1007/978-3-642-20847-8-21
Publication status	Published - 2011
Externally published	Yes
Event	Pacific-Asia Conference on Knowledge Discovery and Data Mining 2011 - Shenzhen, China Duration: 24 May 2011 → 27 May 2011 Conference number: 15th https://link.springer.com/book/10.1007/978-3-642-20841-6 (Proceedings)

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	6635
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Pacific-Asia Conference on Knowledge Discovery and Data Mining 2011
Abbreviated title	PAKDD 2011
Country/Territory	China
City	Shenzhen
Period	24/05/11 → 27/05/11
Internet address	https://link.springer.com/book/10.1007/978-3-642-20841-6 (Proceedings)

Keywords

Novelty detection
one-class classification
spherically shaped boundary
support vector data description

Access to Document

10.1007/978-3-642-20847-8-21

Cite this

Le, T., Tran, D., Ma, W., & Sharma, D. (2011). Multiple distribution data description learning algorithm for novelty detection. In Advances in Knowledge Discovery and Data Mining : 15th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 201, Part II (pp. 246-257). (Lecture Notes in Computer Science ; Vol. 6635). Springer. https://doi.org/10.1007/978-3-642-20847-8-21

@inproceedings{56aea6d99a844fd8ac5bc492550d5257,

title = "Multiple distribution data description learning algorithm for novelty detection",

abstract = "Current data description learning methods for novelty detection such as support vector data description and small sphere with large margin construct a spherically shaped boundary around a normal data set to separate this set from abnormal data. The volume of this sphere is minimized to reduce the chance of accepting abnormal data. However those learning methods do not guarantee that the single spherically shaped boundary can best describe the normal data set if there exist some distinctive data distributions in this set. We propose in this paper a new data description learning method that constructs a set of spherically shaped boundaries to provide a better data description to the normal data set. An optimisation problem is proposed and solving this problem results in an iterative learning algorithm to determine the set of spherically shaped boundaries. We prove that the classification error will be reduced after each iteration in our learning method. Experimental results on 28 well-known data sets show that the proposed method provides lower classification error rates.",

keywords = "Novelty detection, one-class classification, spherically shaped boundary, support vector data description",

author = "Trung Le and Dat Tran and Wanli Ma and Dharmendra Sharma",

year = "2011",

doi = "10.1007/978-3-642-20847-8-21",

language = "English",

isbn = "9783642208461",

series = "Lecture Notes in Computer Science ",

publisher = "Springer",

pages = "246--257",

booktitle = "Advances in Knowledge Discovery and Data Mining",

note = "Pacific-Asia Conference on Knowledge Discovery and Data Mining 2011, PAKDD 2011 ; Conference date: 24-05-2011 Through 27-05-2011",

url = "https://link.springer.com/book/10.1007/978-3-642-20841-6",

}

Le, T, Tran, D, Ma, W & Sharma, D 2011, Multiple distribution data description learning algorithm for novelty detection. in Advances in Knowledge Discovery and Data Mining : 15th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 201, Part II. Lecture Notes in Computer Science , vol. 6635, Springer, pp. 246-257, Pacific-Asia Conference on Knowledge Discovery and Data Mining 2011, Shenzhen, China, 24/05/11. https://doi.org/10.1007/978-3-642-20847-8-21

Multiple distribution data description learning algorithm for novelty detection. / Le, Trung; Tran, Dat; Ma, Wanli et al.
Advances in Knowledge Discovery and Data Mining : 15th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 201, Part II. Springer, 2011. p. 246-257 (Lecture Notes in Computer Science ; Vol. 6635).

Research output: Chapter in Book/Report/Conference proceeding › Conference Paper › Research › peer-review

TY - GEN

T1 - Multiple distribution data description learning algorithm for novelty detection

AU - Le, Trung

AU - Tran, Dat

AU - Ma, Wanli

AU - Sharma, Dharmendra

N1 - Conference code: 15th

PY - 2011

Y1 - 2011

N2 - Current data description learning methods for novelty detection such as support vector data description and small sphere with large margin construct a spherically shaped boundary around a normal data set to separate this set from abnormal data. The volume of this sphere is minimized to reduce the chance of accepting abnormal data. However those learning methods do not guarantee that the single spherically shaped boundary can best describe the normal data set if there exist some distinctive data distributions in this set. We propose in this paper a new data description learning method that constructs a set of spherically shaped boundaries to provide a better data description to the normal data set. An optimisation problem is proposed and solving this problem results in an iterative learning algorithm to determine the set of spherically shaped boundaries. We prove that the classification error will be reduced after each iteration in our learning method. Experimental results on 28 well-known data sets show that the proposed method provides lower classification error rates.

AB - Current data description learning methods for novelty detection such as support vector data description and small sphere with large margin construct a spherically shaped boundary around a normal data set to separate this set from abnormal data. The volume of this sphere is minimized to reduce the chance of accepting abnormal data. However those learning methods do not guarantee that the single spherically shaped boundary can best describe the normal data set if there exist some distinctive data distributions in this set. We propose in this paper a new data description learning method that constructs a set of spherically shaped boundaries to provide a better data description to the normal data set. An optimisation problem is proposed and solving this problem results in an iterative learning algorithm to determine the set of spherically shaped boundaries. We prove that the classification error will be reduced after each iteration in our learning method. Experimental results on 28 well-known data sets show that the proposed method provides lower classification error rates.

KW - Novelty detection

KW - one-class classification

KW - spherically shaped boundary

KW - support vector data description

UR - http://www.scopus.com/inward/record.url?scp=79957956807&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-20847-8-21

DO - 10.1007/978-3-642-20847-8-21

M3 - Conference Paper

AN - SCOPUS:79957956807

SN - 9783642208461

T3 - Lecture Notes in Computer Science

SP - 246

EP - 257

BT - Advances in Knowledge Discovery and Data Mining

PB - Springer

T2 - Pacific-Asia Conference on Knowledge Discovery and Data Mining 2011

Y2 - 24 May 2011 through 27 May 2011

ER -

Multiple distribution data description learning algorithm for novelty detection

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Cite this