Adaptive Anomaly Detection of Coupled Activity Sequences

Publisher:
IEEE
Publication Type:
Journal Article
Citation:
The IEEE Intelligent Informatics Bulletin, 2009, 10 (1), pp. 12 - 16
Issue Date:
2009-01
Full metadata record
Files in This Item:
Filename Description Size
Thumbnail2009001907OK.pdf225.67 kB
Adobe PDF
Many real-life applications often involve multiple sequences, which are coupled with each other. It is unreasonable to either study the multiple coupled sequences separately or simply merge them into one sequence, because the information about their interacting relationships would be lost. Furthermore, such coupled sequences also have frequently significant changes which are likely to degrade the performance of trained model. Taking the detection of abnormal trading activity patterns in stock markets as an example, this paper proposes a Hidden Markov Model-based approach to address the above two issues. Our approach is suitable for sequence analysis on multiple coupled sequences and can adapt to the significant sequence changes automatically. Substantial experiments conducted on a real dataset show that our approach is effective.
Please use this identifier to cite or link to this item: