Exathlon: A Benchmark for Explainable Anomaly Detection over Time Series
This provides a standardized benchmark for researchers and practitioners in time series analytics, enabling more rigorous evaluation and advancement in anomaly detection and explanation methods, though it is incremental as it builds on existing data and techniques.
The authors tackled the lack of community resources for explainable anomaly detection in time series by presenting Exathlon, the first comprehensive public benchmark based on real data from Apache Spark cluster executions, which includes ground truth labels for six anomaly types and supports evaluation of detection and explanation tasks.
Access to high-quality data repositories and benchmarks have been instrumental in advancing the state of the art in many experimental research domains. While advanced analytics tasks over time series data have been gaining lots of attention, lack of such community resources severely limits scientific progress. In this paper, we present Exathlon, the first comprehensive public benchmark for explainable anomaly detection over high-dimensional time series data. Exathlon has been systematically constructed based on real data traces from repeated executions of large-scale stream processing jobs on an Apache Spark cluster. Some of these executions were intentionally disturbed by introducing instances of six different types of anomalous events (e.g., misbehaving inputs, resource contention, process failures). For each of the anomaly instances, ground truth labels for the root cause interval as well as those for the extended effect interval are provided, supporting the development and evaluation of a wide range of anomaly detection (AD) and explanation discovery (ED) tasks. We demonstrate the practical utility of Exathlon's dataset, evaluation methodology, and end-to-end data science pipeline design through an experimental study with three state-of-the-art AD and ED techniques.