Home > Research > Publications & Outputs > OSA
View graph of relations

OSA: one-class recursive SVM algorithm with negative samples for fault detection

Research output: Contribution in Book/Report/Proceedings - With ISBN/ISSNConference contribution/Paperpeer-review

Publication date2013
Host publicationArtificial neural networks and machine learning – ICANN 2013: 23rd International Conference on Artificial Neural Networks Sofia, Bulgaria, September 10-13, 2013. Proceedings
EditorsValeri Mladenov, Petia Koprinkova-Hristova, Günther Palm, Alessandro E. P. Villa, Bruno Appollini, Nikola Kasabov
Place of PublicationBerlin
PublisherSpringer Verlag
Number of pages14
ISBN (Electronic)9783642407284
ISBN (Print)9783642407277
<mark>Original language</mark>English

Publication series

NameLecture Notes in Computer Science
ISSN (Print)0302-9743


In this paper a novel one-class classification approach (called OSA) is proposed. The algorithm is particularly suitable for fault detection in complex technological systems, such as aircraft. This study is based on the capability of one-class support vector machine (SVM) method to classify correctly the observation and measurement data, obtained during the exploitation of the system such as airborne aircraft into a single class of ‘normal’ behavior and, respectively, leave data that is not assigned to this class as suspected anomalies. In order to ensure real time (in flight) application a recursive learning procedure of the method is proposed. The proposed method takes into account both “positive”/“normal” and “negative”/“abnormal” examples of the base class, keeping the overall model structure as an outlier-detection approach. This approach is generic for any fault detection problem (for example in areas such as process control, computer networks, analysis of data from interrogations, etc.). The advantages of the new algorithm based on OSA are verified by comparison with several classifiers, including the traditional one-class SVM. The proposed approach is tested for fault detection problem using real flight data from a large number of aircraft of different make (USA, Western European as well as Russian).