Geenn2026/Time-Series-Library

---
tags:
- time-series
- forecasting
- anomaly-detection
- classification
- TSLib
license: cc-by-4.0
task_categories:
- time-series-forecasting
pretty_name: Time-Series-Library (TSLib)
language:
- en
configs:
- config_name: ETTh1
description: ETT long-term forecasting subset ETTh1 (hourly).
data_files:
- ETT-small/ETTh1.csv
- config_name: ETTh2
description: ETT long-term forecasting subset ETTh2 (hourly).
data_files:
- ETT-small/ETTh2.csv
- config_name: ETTm1
description: ETT long-term forecasting subset ETTm1 (15-min).
data_files:
- ETT-small/ETTm1.csv
- config_name: ETTm2
description: ETT long-term forecasting subset ETTm2 (15-min).
data_files:
- ETT-small/ETTm2.csv
- config_name: electricity
description: Electricity load forecasting (UCI Electricity).
data_files:
- electricity/electricity.csv
- config_name: traffic
description: Traffic volume forecasting.
data_files:
- traffic/traffic.csv
- config_name: weather
description: Weather time-series forecasting.
data_files:
- weather/weather.csv
- config_name: exchange_rate
description: Exchange rate forecasting.
data_files:
- exchange_rate/exchange_rate.csv
- config_name: national_illness
description: Influenza-like illness (ILI) forecasting.
data_files:
- illness/national_illness.csv
- config_name: m4-yearly
description: M4 Yearly forecasting subset.
data_files:
- split: train
path: m4/Yearly-train.csv
- split: test
path: m4/Yearly-test.csv
- config_name: m4-quarterly
description: M4 Quarterly forecasting subset.
data_files:
- split: train
path: m4/Quarterly-train.csv
- split: test
path: m4/Quarterly-test.csv
- config_name: m4-monthly
description: M4 Monthly forecasting subset.
data_files:
- split: train
path: m4/Monthly-train.csv
- split: test
path: m4/Monthly-test.csv
- config_name: m4-weekly
description: M4 Weekly forecasting subset.
data_files:
- split: train
path: m4/Weekly-train.csv
- split: test
path: m4/Weekly-test.csv
- config_name: m4-daily
description: M4 Daily forecasting subset.
data_files:
- split: train
path: m4/Daily-train.csv
- split: test
path: m4/Daily-test.csv
- config_name: m4-hourly
description: M4 Hourly forecasting subset.
data_files:
- split: train
path: m4/Hourly-train.csv
- split: test
path: m4/Hourly-test.csv
- config_name: EthanolConcentration
description: 'UEA multivariate classification: EthanolConcentration.'
data_files:
- split: train
path: EthanolConcentration/EthanolConcentration_TRAIN.ts
- split: test
path: EthanolConcentration/EthanolConcentration_TEST.ts
- config_name: FaceDetection
description: 'UEA multivariate classification: FaceDetection.'
data_files:
- split: train
path: FaceDetection/FaceDetection_TRAIN.ts
- split: test
path: FaceDetection/FaceDetection_TEST.ts
- config_name: Handwriting
description: 'UEA multivariate classification: Handwriting.'
data_files:
- split: train
path: Handwriting/Handwriting_TRAIN.ts
- split: test
path: Handwriting/Handwriting_TEST.ts
- config_name: Heartbeat
description: 'UEA multivariate classification: Heartbeat.'
data_files:
- split: train
path: Heartbeat/Heartbeat_TRAIN.ts
- split: test
path: Heartbeat/Heartbeat_TEST.ts
- config_name: JapaneseVowels
description: 'UEA multivariate classification: JapaneseVowels.'
data_files:
- split: train
path: JapaneseVowels/JapaneseVowels_TRAIN.ts
- split: test
path: JapaneseVowels/JapaneseVowels_TEST.ts
- config_name: PEMS-SF
description: 'UEA multivariate classification: PEMS-SF.'
data_files:
- split: train
path: PEMS-SF/PEMS-SF_TRAIN.ts
- split: test
path: PEMS-SF/PEMS-SF_TEST.ts
- config_name: SelfRegulationSCP1
description: 'UEA multivariate classification: SelfRegulationSCP1.'
data_files:
- split: train
path: SelfRegulationSCP1/SelfRegulationSCP1_TRAIN.ts
- split: test
path: SelfRegulationSCP1/SelfRegulationSCP1_TEST.ts
- config_name: SelfRegulationSCP2
description: 'UEA multivariate classification: SelfRegulationSCP2.'
data_files:
- split: train
path: SelfRegulationSCP2/SelfRegulationSCP2_TRAIN.ts
- split: test
path: SelfRegulationSCP2/SelfRegulationSCP2_TEST.ts
- config_name: SpokenArabicDigits
description: 'UEA multivariate classification: SpokenArabicDigits.'
data_files:
- split: train
path: SpokenArabicDigits/SpokenArabicDigits_TRAIN.ts
- split: test
path: SpokenArabicDigits/SpokenArabicDigits_TEST.ts
- config_name: UWaveGestureLibrary
description: 'UEA multivariate classification: UWaveGestureLibrary.'
data_files:
- split: train
path: UWaveGestureLibrary/UWaveGestureLibrary_TRAIN.ts
- split: test
path: UWaveGestureLibrary/UWaveGestureLibrary_TEST.ts
- config_name: SMD-data
description: Server Machine Dataset (SMD) for anomaly detection — train & test data.
data_files:
- split: train
path: SMD/SMD_train.npy
- split: test
path: SMD/SMD_test.npy
- config_name: SMD-label
description: Server Machine Dataset (SMD) — test anomaly labels.
data_files:
- split: test_label
path: SMD/SMD_test_label.npy
- config_name: MSL-data
description: NASA Mars Science Laboratory (MSL) anomaly detection — train/test arrays.
data_files:
- split: train
path: MSL/MSL_train.npy
- split: test
path: MSL/MSL_test.npy
- config_name: MSL-label
description: MSL anomaly detection — test labels.
data_files:
- split: test_label
path: MSL/MSL_test_label.npy
- config_name: SMAP-data
description: >-
NASA Soil Moisture Active Passive (SMAP) anomaly detection — train/test
arrays.
data_files:
- split: train
path: SMAP/SMAP_train.npy
- split: test
path: SMAP/SMAP_test.npy
- config_name: SMAP-label
description: SMAP anomaly detection — test labels.
data_files:
- split: test_label
path: SMAP/SMAP_test_label.npy
- config_name: PSM-data
description: KPI-based Process/System Monitoring data (train/test).
data_files:
- split: train
path: PSM/train.csv
- split: test
path: PSM/test.csv
- config_name: PSM-label
description: KPI-based Process/System Monitoring labels (test_label).
data_files:
- split: test_label
path: PSM/test_label.csv
- config_name: SWaT
description: Secure Water Treatment (SWaT) anomaly detection, processed data.
data_files:
- split: train
path: SWaT/swat_train2.csv
- split: test
path: SWaT/swat2.csv
size_categories:
- 10M<n<100M
---

# Time-Series-Library (TSLib)

TSLib is an open-source library for deep learning researchers, especially for deep time series analysis.

We provide a neat code base to evaluate advanced deep time series models or develop your model, which covers five mainstream tasks: **long- and short-term forecasting, imputation, anomaly detection, and classification.**

This benchmark collection is designed to evaluate and develop advanced deep time-series models. For an in-depth exploration of current time-series models and their performance, please refer to our paper **[Deep Time Series Models: A Comprehensive Survey and Benchmark](https://arxiv.org/abs/2407.13278)**.

To get started with the codebase and contribute, please visit the **[GitHub repository](https://github.com/thuml/Time-Series-Library)**.

## Dataset Overview

| **Tasks** | **Benchmarks** | **Metrics** | **Series Length** |
|-------------------|-------------------------------------------------------------------------------|--------------------------------------|-----------------------|
| **Forecasting** | **Long-term:** ETT (4 subsets), Electricity, Traffic, Weather, Exchange, ILI | MSE, MAE | 96\~720 (ILI: 24\~60) |
| | **Short-term:** M4 (6 subsets) | SMAPE, MASE, OWA | 6\~48 |
| **Imputation** | ETT (4 subsets), Electricity, Weather | MSE, MAE | 96 |
| **Classification** | UEA (10 subsets) | Accuracy | 29\~1751 |
| **Anomaly Detection** | SMD, MSL, SMAP, SWaT, PSM | Precision, Recall, F1-Score | 100 |

## File Structure
```
Time-Series-Library/
├── ETT-small/
├── EthanolConcentration/
├── FaceDetection/
├── Handwriting/
├── Heartbeat/
├── JapaneseVowels/
├── MSL/
├── PEMS-SF/
├── PSM/
├── SMAP/
├── SMD/
├── SWaT/
├── SelfRegulationSCP1/
├── SelfRegulationSCP2/
├── SpokenArabicDigits/
├── UWaveGestureLibrary/
├── electricity/
├── exchange_rate/
├── illness/
├── m4/
├── traffic/
├── weather/
├── .gitattributes
└── README.md
```

## Usage

You can load the dataset directly using the `datasets` library:

```
from datasets import load_dataset
dataset = load_dataset("thuml/Time-Series-Library", "ETTh1")
```

Or download specific files with hf_hub_download:

```
from huggingface_hub import hf_hub_download
hf_hub_download("thuml/Time-Series-Library", "ETT-small/ETTh1.csv", repo_type="dataset")
```

## License
This dataset is released under the CC BY 4.0 License.

## Citation

If you find this repo useful, please cite our paper.

```
@inproceedings{wu2023timesnet,
title={TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis},
author={Haixu Wu and Tengge Hu and Yong Liu and Hang Zhou and Jianmin Wang and Mingsheng Long},
booktitle={International Conference on Learning Representations},
year={2023},
}

@article{wang2024tssurvey,
title={Deep Time Series Models: A Comprehensive Survey and Benchmark},
author={Yuxuan Wang and Haixu Wu and Jiaxiang Dong and Yong Liu and Mingsheng Long and Jianmin Wang},
booktitle={arXiv preprint arXiv:2407.13278},
year={2024},
}
```

tags:
- 时间序列（time-series）
- 预测（forecasting）
- 异常检测（anomaly-detection）
- 分类（classification）
- 时间序列库（Time-Series-Library, TSLib）
license: cc-by-4.0
task_categories:
- 时间序列预测（time-series-forecasting）
pretty_name: 时间序列库（Time-Series-Library, TSLib）
language:
- en
configs:
- config_name: ETTh1
description: ETT长期预测子集ETTh1（小时级数据）。
data_files:
- ETT-small/ETTh1.csv
- config_name: ETTh2
description: ETT长期预测子集ETTh2（小时级数据）。
data_files:
- ETT-small/ETTh2.csv
- config_name: ETTm1
description: ETT长期预测子集ETTm1（15分钟采样）。
data_files:
- ETT-small/ETTm1.csv
- config_name: ETTm2
description: ETT长期预测子集ETTm2（15分钟采样）。
data_files:
- ETT-small/ETTm2.csv
- config_name: electricity
description: 电力负荷预测（UCI电力数据集）。
data_files:
- electricity/electricity.csv
- config_name: traffic
description: 交通流量预测。
data_files:
- traffic/traffic.csv
- config_name: weather
description: 气象时间序列预测。
data_files:
- weather/weather.csv
- config_name: exchange_rate
description: 汇率预测。
data_files:
- exchange_rate/exchange_rate.csv
- config_name: national_illness
description: 流感样病例（ILI）预测。
data_files:
- illness/national_illness.csv
- config_name: m4-yearly
description: M4年度预测子集。
data_files:
- split: train
path: m4/Yearly-train.csv
- split: test
path: m4/Yearly-test.csv
- config_name: m4-quarterly
description: M4季度预测子集。
data_files:
- split: train
path: m4/Quarterly-train.csv
- split: test
path: m4/Quarterly-test.csv
- config_name: m4-monthly
description: M4月度预测子集。
data_files:
- split: train
path: m4/Monthly-train.csv
- split: test
path: m4/Monthly-test.csv
- config_name: m4-weekly
description: M4周度预测子集。
data_files:
- split: train
path: m4/Weekly-train.csv
- split: test
path: m4/Weekly-test.csv
- config_name: m4-daily
description: M4日度预测子集。
data_files:
- split: train
path: m4/Daily-train.csv
- split: test
path: m4/Daily-test.csv
- config_name: m4-hourly
description: M4小时级预测子集。
data_files:
- split: train
path: m4/Hourly-train.csv
- split: test
path: m4/Hourly-test.csv
- config_name: EthanolConcentration
description: UEA多变量分类任务：EthanolConcentration。
data_files:
- split: train
path: EthanolConcentration/EthanolConcentration_TRAIN.ts
- split: test
path: EthanolConcentration/EthanolConcentration_TEST.ts
- config_name: FaceDetection
description: UEA多变量分类任务：FaceDetection。
data_files:
- split: train
path: FaceDetection/FaceDetection_TRAIN.ts
- split: test
path: FaceDetection/FaceDetection_TEST.ts
- config_name: Handwriting
description: UEA多变量分类任务：Handwriting。
data_files:
- split: train
path: Handwriting/Handwriting_TRAIN.ts
- split: test
path: Handwriting/Handwriting_TEST.ts
- config_name: Heartbeat
description: UEA多变量分类任务：Heartbeat。
data_files:
- split: train
path: Heartbeat/Heartbeat_TRAIN.ts
- split: test
path: Heartbeat/Heartbeat_TEST.ts
- config_name: JapaneseVowels
description: UEA多变量分类任务：JapaneseVowels。
data_files:
- split: train
path: JapaneseVowels/JapaneseVowels_TRAIN.ts
- split: test
path: JapaneseVowels/JapaneseVowels_TEST.ts
- config_name: PEMS-SF
description: UEA多变量分类任务：PEMS-SF。
data_files:
- split: train
path: PEMS-SF/PEMS-SF_TRAIN.ts
- split: test
path: PEMS-SF/PEMS-SF_TEST.ts
- config_name: SelfRegulationSCP1
description: UEA多变量分类任务：SelfRegulationSCP1。
data_files:
- split: train
path: SelfRegulationSCP1/SelfRegulationSCP1_TRAIN.ts
- split: test
path: SelfRegulationSCP1/SelfRegulationSCP1_TEST.ts
- config_name: SelfRegulationSCP2
description: UEA多变量分类任务：SelfRegulationSCP2。
data_files:
- split: train
path: SelfRegulationSCP2/SelfRegulationSCP2_TRAIN.ts
- split: test
path: SelfRegulationSCP2/SelfRegulationSCP2_TEST.ts
- config_name: SpokenArabicDigits
description: UEA多变量分类任务：SpokenArabicDigits。
data_files:
- split: train
path: SpokenArabicDigits/SpokenArabicDigits_TRAIN.ts
- split: test
path: SpokenArabicDigits/SpokenArabicDigits_TEST.ts
- config_name: UWaveGestureLibrary
description: UEA多变量分类任务：UWaveGestureLibrary。
data_files:
- split: train
path: UWaveGestureLibrary/UWaveGestureLibrary_TRAIN.ts
- split: test
path: UWaveGestureLibrary/UWaveGestureLibrary_TEST.ts
- config_name: SMD-data
description: 服务器机器数据集（SMD）异常检测——训练与测试数据。
data_files:
- split: train
path: SMD/SMD_train.npy
- split: test
path: SMD/SMD_test.npy
- config_name: SMD-label
description: 服务器机器数据集（SMD）——测试集异常标签。
data_files:
- split: test_label
path: SMD/SMD_test_label.npy
- config_name: MSL-data
description: 美国国家航空航天局火星科学实验室（MSL）异常检测——训练/测试数组。
data_files:
- split: train
path: MSL/MSL_train.npy
- split: test
path: MSL/MSL_test.npy
- config_name: MSL-label
description: MSL异常检测——测试集标签。
data_files:
- split: test_label
path: MSL/MSL_test_label.npy
- config_name: SMAP-data
description: 美国国家航空航天局土壤湿度主动被动卫星（SMAP）异常检测——训练/测试数组。
data_files:
- split: train
path: SMAP/SMAP_train.npy
- split: test
path: SMAP/SMAP_test.npy
- config_name: SMAP-label
description: SMAP异常检测——测试集标签。
data_files:
- split: test_label
path: SMAP/SMAP_test_label.npy
- config_name: PSM-data
description: 基于KPI的流程/系统监测数据（训练/测试集）。
data_files:
- split: train
path: PSM/train.csv
- split: test
path: PSM/test.csv
- config_name: PSM-label
description: 基于KPI的流程/系统监测标签（测试集标签）。
data_files:
- split: test_label
path: PSM/test_label.csv
- config_name: SWaT
description: 安全水处理（SWaT）异常检测，预处理数据。
data_files:
- split: train
path: SWaT/swat_train2.csv
- split: test
path: SWaT/swat2.csv
size_categories:
- 10M<n<100M

# 时间序列库（Time-Series-Library, TSLib）

时间序列库（TSLib）是面向深度学习研究者的开源工具库，尤其聚焦于深度时间序列分析领域。

我们提供了简洁规范的代码框架，用于评估先进的深度时间序列模型或开发自定义模型，覆盖五大主流任务：**长短期预测、缺失值填充、异常检测与分类**。

本基准数据集集合专为评估与开发先进深度时间序列模型而设计。若需深入了解当前主流时间序列模型及其性能表现，请参阅我们的论文**《深度时间序列模型：全面综述与基准测试》（Deep Time Series Models: A Comprehensive Survey and Benchmark，arXiv:2407.13278）**。

若需上手使用本代码库或参与贡献，请访问**[GitHub仓库](https://github.com/thuml/Time-Series-Library)**。

## 数据集概览

| **任务类型** | **基准数据集** | **评价指标** | **序列长度** |
|-------------------|-------------------------------------------------------------------------------|--------------------------------------|-----------------------|
| **预测** | **长期预测：** ETT（4个子集）、电力负荷、交通流量、气象数据、汇率、流感样病例（ILI） | 均方误差（MSE）、平均绝对误差（MAE） | 96~720（ILI：24~60） |
| | **短期预测：** M4（6个子集） | 对称平均绝对百分比误差（SMAPE）、平均绝对标度误差（MASE）、最优加权平均（OWA） | 6~48 |
| **缺失值填充** | ETT（4个子集）、电力负荷、气象数据 | MSE、MAE | 96 |
| **分类** | UEA（10个子集） | 准确率（Accuracy） | 29~1751 |
| **异常检测** | SMD、MSL、SMAP、SWaT、PSM | 精确率（Precision）、召回率（Recall）、F1值（F1-Score） | 100 |

## 文件结构

Time-Series-Library/
├── ETT-small/
├── EthanolConcentration/
├── FaceDetection/
├── Handwriting/
├── Heartbeat/
├── JapaneseVowels/
├── MSL/
├── PEMS-SF/
├── PSM/
├── SMAP/
├── SMD/
├── SWaT/
├── SelfRegulationSCP1/
├── SelfRegulationSCP2/
├── SpokenArabicDigits/
├── UWaveGestureLibrary/
├── electricity/
├── exchange_rate/
├── illness/
├── m4/
├── traffic/
├── weather/
├── .gitattributes
└── README.md

## 使用方法

你可以直接通过`datasets`库加载本数据集：
python
from datasets import load_dataset
dataset = load_dataset("thuml/Time-Series-Library", "ETTh1")

或者使用`hf_hub_download`下载指定文件：
python
from huggingface_hub import hf_hub_download
hf_hub_download("thuml/Time-Series-Library", "ETT-small/ETTh1.csv", repo_type="dataset")

## 许可证
本数据集采用CC BY 4.0许可证发布。

## 引用

若你认为本仓库对你的研究有所帮助，请引用以下论文：

bibtex
@inproceedings{wu2023timesnet,
title={TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysis},
author={Haixu Wu and Tengge Hu and Yong Liu and Hang Zhou and Jianmin Wang and Mingsheng Long},
booktitle={International Conference on Learning Representations},
year={2023},
}

@article{wang2024tssurvey,
title={Deep Time Series Models: A Comprehensive Survey and Benchmark},
author={Yuxuan Wang and Haixu Wu and Jiaxiang Dong and Yong Liu and Mingsheng Long and Jianmin Wang},
booktitle={arXiv preprint arXiv:2407.13278},
year={2024},
}