Powertrain Anomaly Time series bencHmark (PATH) dataset
收藏资源简介:
PATH数据集是由莱顿大学创建的一个用于评估在线无监督异常检测方法的多变量时间序列数据集。该数据集通过先进的仿真工具生成,反映了汽车动力系统的真实行为,包括其多变量、动态和变状态属性。数据集旨在支持无监督和半监督异常检测设置,以及时间序列生成和预测任务。数据集的创建过程基于物理启发的仿真模型,特别是电动车辆的仿真模型。PATH数据集主要应用于工业过程的数字化和系统行为建模,旨在解决时间序列中的异常检测问题。
The PATH dataset is a multivariate time series dataset created by Leiden University for evaluating online unsupervised anomaly detection methods. Generated using advanced simulation tools, it captures the real-world behavior of automotive powertrain systems, including its multivariate, dynamic, and state-varying properties. This dataset is designed to support unsupervised and semi-supervised anomaly detection settings, as well as time series generation and forecasting tasks. Its creation process is based on physics-informed simulation models, particularly those for electric vehicles. The PATH dataset is primarily applied to the digitization of industrial processes and system behavior modeling, aiming to address anomaly detection problems in time series.




