ML Anthology

Parameterless Semi-Supervised Anomaly Detection in Univariate Time Series

Oleg Iegorov, Sebastian Fischmeister
ECML-PKDD 2020 pp. 644-659
doi:10.1007/978-3-030-67658-2_37 /ecmlpkdd/2020/iegorov2020ecmlpkdd-parameterless/

Abstract

Anomaly detection algorithms that operate without human intervention are needed when dealing with large time series data coming from poorly understood processes. At the same time, common techniques expect the user to provide precise information about the data generating process or to manually tune various parameters. We present SIM-AD: a semi-supervised approach to detecting anomalies in univariate time series data that operates without any user-defined parameters. The approach involves converting time series using our proposed Sojourn Time Representation and then applying modal clustering-based anomaly detection on the converted data. We evaluate SIM-AD on three publicly available time series datasets from different domains and compare its accuracy to the PAV and RRA anomaly detection algorithms. We conclude that SIM-AD outperforms the evaluated approaches with respect to accuracy on trendless time series data.

PDF ECML-PKDD Semantic Scholar

Cite

Text

Iegorov and Fischmeister. "Parameterless Semi-Supervised Anomaly Detection in Univariate Time Series." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2020. doi:10.1007/978-3-030-67658-2_37

Markdown

[Iegorov and Fischmeister. "Parameterless Semi-Supervised Anomaly Detection in Univariate Time Series." European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, 2020.](https://mlanthology.org/ecmlpkdd/2020/iegorov2020ecmlpkdd-parameterless/) doi:10.1007/978-3-030-67658-2_37

BibTeX

@inproceedings{iegorov2020ecmlpkdd-parameterless,
  title     = {{Parameterless Semi-Supervised Anomaly Detection in Univariate Time Series}},
  author    = {Iegorov, Oleg and Fischmeister, Sebastian},
  booktitle = {European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases},
  year      = {2020},
  pages     = {644-659},
  doi       = {10.1007/978-3-030-67658-2_37},
  url       = {https://mlanthology.org/ecmlpkdd/2020/iegorov2020ecmlpkdd-parameterless/}
}