Flow-Bench

Flow-Bench是一个用于计算工作流异常检测的数据集，由南加州大学等机构创建。该数据集包含超过1200个工作流执行的原始日志，涵盖正常和异常条件，旨在通过机器学习技术提高工作流执行的可靠性。数据集通过系统地注入异常并收集分布式基础设施上的工作流执行日志来构建。Flow-Bench不仅包括新的工作流数据，还提供了对现有开放数据集的统计总结和深入分析，以及对无监督异常检测技术的基准测试。该数据集适用于科学、工程和数据科学研究，特别是在处理复杂现象的模拟和建模方面。

Flow-Bench is a dataset for computational workflow anomaly detection, developed by institutions including the University of Southern California and other relevant organizations. This dataset contains over 1200 raw logs of workflow executions, covering both normal and anomalous conditions, with the objective of improving the reliability of workflow execution via machine learning technologies. It was constructed by systematically injecting anomalies and collecting workflow execution logs across distributed infrastructures. Flow-Bench not only includes newly created workflow data, but also provides statistical summaries and in-depth analyses of existing open datasets, as well as benchmark evaluations for unsupervised anomaly detection techniques. This dataset is applicable to scientific, engineering, and data science research, particularly in the simulation and modeling of complex phenomena.