Unsupervised parameter estimation for one-class support vector machines

Ghafoori, Zahra, Rajasegarar, Sutharshan, Erfani, Sarah M., Karunasekera, Shanika and Leckie, Christopher A. 2016, Unsupervised parameter estimation for one-class support vector machines. In Bailey, James, Khan, Latifur, Washio, Takashi, Dobbie, Gill, Zhexue Huang, Joshua and Wang, Ruili (ed), Advances in knowledge discovery and data mining, Springer, Berlin, Germany, pp.183-195, doi: 10.1007/978-3-319-31750-2_15.

Attached Files
Name Description MIMEType Size Downloads

Title Unsupervised parameter estimation for one-class support vector machines
Author(s) Ghafoori, Zahra
Rajasegarar, SutharshanORCID iD for Rajasegarar, Sutharshan orcid.org/0000-0002-6559-6736
Erfani, Sarah M.
Karunasekera, Shanika
Leckie, Christopher A.
Title of book Advances in knowledge discovery and data mining
Editor(s) Bailey, James
Khan, Latifur
Washio, Takashi
Dobbie, Gill
Zhexue Huang, Joshua
Wang, Ruili
Publication date 2016
Series Lecture Notes in Computer Science
Chapter number 15
Total chapters 44
Start page 183
End page 195
Total pages 13
Publisher Springer
Place of Publication Berlin, Germany
Keyword(s) one-class support vector machine
support vector data description
outlier detection
parameter estimation
Summary Although the hyper-plane based One-Class Support Vector Machine (OCSVM) and the hyper-spherical based Support Vector Data Description (SVDD) algorithms have been shown to be very effective in detecting outliers, their performance on noisy and unlabeled training data has not been widely studied. Moreover, only a few heuristic approaches have been proposed to set the different parameters of these methods in an unsupervised manner. In this paper, we propose two unsupervised methods for estimating the optimal parameter settings to train OCSVM and SVDD models, based on analysing the structure of the data. We show that our heuristic is substantially faster than existing parameter estimation approaches while its accuracy is comparable with supervised parameter learning methods, such as grid-search with crossvalidation on labeled data. In addition, our proposed approaches can be used to prepare a labeled data set for a OCSVM or a SVDD from unlabeled data.
Notes This publication is included within Part II of the 20th PAKDD 2016 Pacific-Asia Conference held in Auckland, New Zealand, 19-22 April.
ISBN 9783319317496
ISSN 0302-9743
Language eng
DOI 10.1007/978-3-319-31750-2_15
Field of Research 080109 Pattern Recognition and Data Mining
Socio Economic Objective 899999 Information and Communication Services not elsewhere classified
HERDC Research category B1 Book chapter
ERA Research output type B Book chapter
Copyright notice ©2016, Springer
Persistent URL http://hdl.handle.net/10536/DRO/DU:30086387

Document type: Book Chapter
Collection: School of Information Technology
Connect to link resolver
Unless expressly stated otherwise, the copyright for items in DRO is owned by the author, with all rights reserved.

Version Filter Type
Citation counts: TR Web of Science Citation Count  Cited 1 times in TR Web of Science
Scopus Citation Count Cited 4 times in Scopus
Google Scholar Search Google Scholar
Access Statistics: 168 Abstract Views, 7 File Downloads  -  Detailed Statistics
Created: Thu, 29 Sep 2016, 11:01:57 EST

Every reasonable effort has been made to ensure that permission has been obtained for items included in DRO. If you believe that your rights have been infringed by this repository, please contact drosupport@deakin.edu.au.