UCR time series classification archive
收藏资源简介:
UCR时间序列分类档案是一个包含128个异构单变量时间序列数据集的存储库,来自于各种应用领域,其中112个数据集具有相等的序列长度。该数据集被广泛用于时间序列分类方法的研究和评估。文中指出,当前基准数据集中有许多数据集的表格特征较为明显,或者已经进行了良好的分段,这使得时间信息在分类中的重要性降低。为此,作者提出了UCR Augmented基准,通过向数据集中引入失配来减少表格特征的影响,强调时间信息的重要性。
The UCR Time Series Classification Archive is a repository containing 128 heterogeneous univariate time series datasets from diverse application domains, among which 112 datasets have uniform sequence lengths. This dataset collection has been widely utilized for the research and evaluation of time series classification methods. It is noted that many datasets in current benchmark datasets exhibit prominent tabular-like features or have been properly segmented, which reduces the importance of temporal information in classification tasks. To address this issue, the authors proposed the UCR Augmented benchmark, which introduces mismatches into the datasets to mitigate the impact of tabular-like features and emphasize the significance of temporal information.

- 1Revisit Time Series Classification Benchmark: The Impact of Temporal Information for Classification新南威尔士大学 · 2025年



