Proceedings of the

The 33rd European Safety and Reliability Conference (ESREL 2023)
3 – 8 September 2023, Southampton, UK

Interpretation of Influential Factors for AI-Based Anomaly Detection

Sheng Dinga, Adrian Wolfb and Andrey Morozovc

Institute for Automation Technology and Software System, University of Stuttgart, Germany.


Anomaly detection is a crucial task in a wide range of industries and domains. The ability to identify abnormal patterns and behaviors in time series data can help detect potential issues, prevent downtime, reduce maintenance costs, and improve the overall performance of systems and processes. This paper focuses on analyzing the significant factors that affect the accuracy and dependability of AI-based time series anomaly detection. The objective is to provide comprehensive insights into interpreting these factors and to explore their impact on the performance. Our study's outcomes shall assist researchers and practitioners in selecting the most appropriate approaches for anomaly detection tasks in diverse domains.

Keywords: Fault detection, Anomaly detection, Machine learning, Deep learning.

Download PDF