Hidden Markov Anomaly Detection

Nico Goernitz, Mikio Braun, Marius Kloft
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:1833-1842, 2015.

Abstract

We introduce a new anomaly detection methodology for data with latent dependency structure. As a particular instantiation, we derive a hidden Markov anomaly detector that extends the regular one-class support vector machine. We optimize the approach, which is non-convex, via a DC (difference of convex functions) algorithm, and show that the parameter v can be conveniently used to control the number of outliers in the model. The empirical evaluation on artificial and real data from the domains of computational biology and computational sustainability shows that the approach can achieve significantly higher anomaly detection performance than the regular one-class SVM.

Cite this Paper

BibTeX
@InProceedings{pmlr-v37-goernitz15, title = {Hidden Markov Anomaly Detection}, author = {Goernitz, Nico and Braun, Mikio and Kloft, Marius}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {1833--1842}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/goernitz15.pdf}, url = {https://proceedings.mlr.press/v37/goernitz15.html}, abstract = {We introduce a new anomaly detection methodology for data with latent dependency structure. As a particular instantiation, we derive a hidden Markov anomaly detector that extends the regular one-class support vector machine. We optimize the approach, which is non-convex, via a DC (difference of convex functions) algorithm, and show that the parameter v can be conveniently used to control the number of outliers in the model. The empirical evaluation on artificial and real data from the domains of computational biology and computational sustainability shows that the approach can achieve significantly higher anomaly detection performance than the regular one-class SVM.} }
Endnote
%0 Conference Paper %T Hidden Markov Anomaly Detection %A Nico Goernitz %A Mikio Braun %A Marius Kloft %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-goernitz15 %I PMLR %P 1833--1842 %U https://proceedings.mlr.press/v37/goernitz15.html %V 37 %X We introduce a new anomaly detection methodology for data with latent dependency structure. As a particular instantiation, we derive a hidden Markov anomaly detector that extends the regular one-class support vector machine. We optimize the approach, which is non-convex, via a DC (difference of convex functions) algorithm, and show that the parameter v can be conveniently used to control the number of outliers in the model. The empirical evaluation on artificial and real data from the domains of computational biology and computational sustainability shows that the approach can achieve significantly higher anomaly detection performance than the regular one-class SVM.
RIS
TY - CPAPER TI - Hidden Markov Anomaly Detection AU - Nico Goernitz AU - Mikio Braun AU - Marius Kloft BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-goernitz15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 1833 EP - 1842 L1 - http://proceedings.mlr.press/v37/goernitz15.pdf UR - https://proceedings.mlr.press/v37/goernitz15.html AB - We introduce a new anomaly detection methodology for data with latent dependency structure. As a particular instantiation, we derive a hidden Markov anomaly detector that extends the regular one-class support vector machine. We optimize the approach, which is non-convex, via a DC (difference of convex functions) algorithm, and show that the parameter v can be conveniently used to control the number of outliers in the model. The empirical evaluation on artificial and real data from the domains of computational biology and computational sustainability shows that the approach can achieve significantly higher anomaly detection performance than the regular one-class SVM. ER -
APA
Goernitz, N., Braun, M. & Kloft, M.. (2015). Hidden Markov Anomaly Detection. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:1833-1842 Available from https://proceedings.mlr.press/v37/goernitz15.html.

Related Material