OXI: An online tool for visualization and annotation of satellite time series data

Article Chemistry, Multidisciplinary

Is It Worth It? Comparing Six Deep and Classical Methods for Unsupervised Anomaly Detection in Time Series

Ferdinand Rewicki et al.

Summary: Detecting anomalies in time series data is crucial in various fields. In this study, we compared six unsupervised anomaly detection methods of varying complexity to determine their performance and suitability for different types of anomalies. The evaluation was done using the UCR anomaly archive, and the methods were adjusted with appropriate hyperparameters. Our experiments reveal that classical machine learning methods generally outperform deep learning methods in detecting a range of anomaly types.

APPLIED SCIENCES-BASEL (2023)

Add to Collection

Article Computer Science, Artificial Intelligence

Deep learning for anomaly detection in multivariate time series: Approaches, applications, and challenges

Gen Li et al.

Summary: This study reviews the state-of-the-art deep learning techniques for anomaly detection in different types, including abnormal time points, time intervals, and time series. Long short-term memory and autoencoders are commonly used methods, and dynamic graphs have been implemented to detect abnormal time intervals. However, anomaly detection still faces challenges in explaining the anomalies.

INFORMATION FUSION (2023)

Add to Collection

Article Engineering, Aerospace

Evaluating algorithms for anomaly detection in satellite telemetry data

Jakub Nalepa et al.

Summary: Detecting anomalies in spacecraft telemetry data is crucial for ensuring safe operation. This paper presents a two-stage technique for anomalous event detection and introduces new quality metrics to address the shortcomings of existing metrics in capturing temporal aspects. Experimental study reveals that conclusions based solely on overlap metrics may lead to biased observations.

ACTA ASTRONAUTICA (2022)

Add to Collection

Article Computer Science, Information Systems

Anomaly Detection in Time Series: A Comprehensive Evaluation

Sebastian Schmidl et al.

Summary: Detecting anomalous subsequences in time series data is crucial for various fields like manufacturing processes, finance applications, and health care monitoring. An increase in the number and variety of anomaly detection algorithms has not been systematically evaluated in the past. This study carefully evaluates state-of-the-art anomaly detection algorithms to address the challenge of selecting the best detection technique for a given task.

PROCEEDINGS OF THE VLDB ENDOWMENT (2022)

Add to Collection

Article Engineering, Electrical & Electronic

Improved deep learning based telemetry data anomaly detection to enhance spacecraft operation reliability

L. Yang et al.

Summary: Spacecraft is a complex system integrating a large number of electronic components and payloads, and abnormal events often occur during in-orbit operation. This paper proposes an improved deep learning based anomaly detection method, combining the highly nonlinear modeling ability of LSTM networks with multi-scale anomaly detection strategy to increase detection performance. The effectiveness of the proposed method is verified using NASA benchmark spacecraft data and Beidou Navigation Satellite hydrogen clock data.

MICROELECTRONICS RELIABILITY (2021)

Add to Collection

Article Computer Science, Information Systems

Exathlon: A Benchmark for Explainable Anomaly Detection over Time Series

Vincent Jacob et al.

Summary: Access to high-quality data repositories and benchmarks is crucial for advancing experimental research. Exathlon is introduced as the first comprehensive public benchmark for explainable anomaly detection over high-dimensional time series data, systematically constructed based on real data traces from repeated executions of large-scale stream processing jobs on an Apache Spark cluster.

PROCEEDINGS OF THE VLDB ENDOWMENT (2021)

Add to Collection

Article Engineering, Electrical & Electronic