Window Size Effects in Time Series Anomaly Detection: An Evaluation of Pitchy Anomaly Detection

Ekin Can Erkuş () and Vilda Purutçuoğlu ()
Additional contact information
Ekin Can Erkuş: Turkey R&D Center, Intelligent Application DC, Huawei Technologies Co. Ltd
Vilda Purutçuoğlu: Middle East Technical University, Department of Statistics

A chapter in Directional and Multivariate Statistics, 2025, pp 349-361 from Springer

Abstract: Abstract Real-time anomaly detection in time series data is crucial for various applications and requires several parameters to be defined prior to the proper analyses. The window size is one of those parameters which indicates the batch data size that is analyzed in each iteration. A recently published Pitchy Anomaly Detection (PAD) algorithm, which combines spectral and time-domain analysis, is used to investigate the effects of window size on anomaly detection performance and computational complexities. The study uses two pre-annotated ECG datasets, and a synthetically generated dataset under different data size conditions to perform the experiments. Results show that smaller window sizes generally lead to improved accuracy and F1-scores, especially in detecting ventricular anomalies. However, excessively small windows can lead to a loss of contextual information and negatively impact overall performance.

Date: 2025
References: Add references at CitEc
Citations:

There are no downloads for this item, see the EconPapers FAQ for hints about obtaining it.

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:spr:sprchp:978-981-96-2004-3_18

Ordering information: This item can be ordered from
http://www.springer.com/9789819620043

DOI: 10.1007/978-981-96-2004-3_18

Access Statistics for this chapter

More chapters in Springer Books from Springer
Bibliographic data for series maintained by Sonal Shukla () and Springer Nature Abstracting and Indexing ().