TSGBench
收藏资源简介:
TSGBench是由新加坡国立大学研究团队创建的时间序列生成基准数据集,旨在为时间序列生成方法提供一个统一和全面的评估平台。该数据集包含10个公开可用的真实世界数据集,涵盖多个应用领域,如金融、能源、医疗等。数据集通过标准化预处理流程确保数据质量,并提供了一套全面的评估指标,包括模型基础、特征基础、距离基础等,以评估时间序列生成方法的性能。此外,TSGBench还引入了一种基于域适应的泛化测试,以评估方法在不同数据域间的适应能力。
TSGBench is a time series generation benchmark dataset created by the research team from the National University of Singapore, which aims to provide a unified and comprehensive evaluation platform for time series generation methods. This dataset comprises 10 publicly available real-world datasets spanning multiple application domains, including finance, energy, healthcare and others. It adopts a standardized preprocessing pipeline to guarantee data quality, and offers a comprehensive suite of evaluation metrics covering model-based, feature-based and distance-based categories to assess the performance of time series generation methods. Furthermore, TSGBench introduces a domain adaptation-based generalization test to evaluate the cross-domain adaptation ability of the evaluated methods.




