UCR Time Series Anomaly Archive
收藏资源简介:
UCR Time Series Anomaly Archive是由加州大学河滨分校计算机科学与工程系的Renjie Wu和Eamonn J. Keogh创建的一个新的时间序列异常检测基准数据集。该数据集旨在解决现有基准数据集存在的多种缺陷,如简单性、不切实际的异常密度、错误的标注和运行至故障偏差等问题。数据集涵盖多个领域,包括医学、体育、昆虫学等,旨在为研究者提供一个更可靠的比较和评估异常检测算法的平台。
The UCR Time Series Anomaly Archive is a novel benchmark dataset for time series anomaly detection, created by Renjie Wu and Eamonn J. Keogh from the Department of Computer Science and Engineering, University of California, Riverside. This dataset is designed to address multiple limitations of existing benchmark datasets, including oversimplification, unrealistic anomaly density, incorrect annotations, and run-to-failure bias. Spanning multiple domains such as medicine, sports, and entomology, it aims to provide researchers with a more reliable platform for comparing and evaluating anomaly detection algorithms.

- 1Current Time Series Anomaly Detection Benchmarks are Flawed and are Creating the Illusion of Progress加州大学河滨分校计算机科学与工程系 · 2022年



