source

​
Weather

 Weather (freq:str='10M', name:str='weather', n_ts:int=21,
          test_size:int=10539, val_size:int=5270, horizons:Tuple[int]=(96,
          192, 336, 720))

This Weather dataset contains the 2020 year of 21 meteorological measurements recorded every 10 minutes from the Weather Station of the Max Planck Biogeochemistry Institute in Jena, Germany.

Reference: Wu, H., Xu, J., Wang, J., and Long, M. Autoformer: Decomposition Transformers with auto-correlation for long-term series forecasting. NeurIPS 2021. https://arxiv.org/abs/2106.13008.

source

​
TrafficL

 TrafficL (freq:str='H', name:str='traffic', n_ts:int=862,
           test_size:int=3508, val_size:int=1756, horizons:Tuple[int]=(96,
           192, 336, 720))

This large Traffic dataset was collected by the California Department of Transportation, it reports road hourly occupancy rates of 862 sensors, from January 2015 to December 2016.

Reference: Lai, G., Chang, W., Yang, Y., and Liu, H. Modeling Long and Short-Term Temporal Patterns with Deep Neural Networks. SIGIR 2018. http://arxiv.org/abs/1703.07015.

Wu, H., Xu, J., Wang, J., and Long, M. Autoformer: Decomposition Transformers with auto-correlation for long-term series forecasting. NeurIPS 2021. https://arxiv.org/abs/2106.13008.

source

​
ECL

 ECL (freq:str='15T', name:str='ECL', n_ts:int=321, n_time:int=26304,
      test_size:int=5260, val_size:int=2632, horizons:Tuple[int]=(96, 192,
      336, 720))

The Electricity dataset reports the fifteen minute electricity consumption (KWh) of 321 customers from 2012 to 2014. For comparability, we aggregate it hourly.

Reference: Li, S et al. Enhancing the locality and breaking the memory bottleneck of Transformer on time series forecasting. NeurIPS 2019. http://arxiv.org/abs/1907.00235.

source

​
ETTm2

 ETTm2 (freq:str='15T', name:str='ETTm2', n_ts:int=7, n_time:int=57600,
        test_size:int=11520, val_size:int=11520, horizons:Tuple[int]=(96,
        192, 336, 720))

The ETTm2 dataset monitors an electricity transformer from a region of a province of China including oil temperature and variants of load (such as high useful load and high useless load) from July 2016 to July 2018 at a fifteen minute frequency.

Reference: Zhou, et al. Informer: Beyond Efficient Transformer for Long Sequence Time-Series Forecasting. AAAI 2021. https://arxiv.org/abs/2012.07436

source

​
ETTm1

 ETTm1 (freq:str='15T', name:str='ETTm1', n_ts:int=7, n_time:int=57600,
        test_size:int=11520, val_size:int=11520, horizons:Tuple[int]=(96,
        192, 336, 720))

The ETTm1 dataset monitors an electricity transformer from a region of a province of China including oil temperature and variants of load (such as high useful load and high useless load) from July 2016 to July 2018 at a fifteen minute frequency.

source

​
ETTh2

 ETTh2 (freq:str='H', name:str='ETTh2', n_ts:int=7, n_time:int=14400,
        test_size:int=2880, val_size:int=2880, horizons:Tuple[int]=(96,
        192, 336, 720))

The ETTh2 dataset monitors an electricity transformer from a region of a province of China including oil temperature and variants of load (such as high useful load and high useless load) from July 2016 to July 2018 at an hourly frequency.

source

​
ETTh1

 ETTh1 (freq:str='H', name:str='ETTh1', n_ts:int=7, n_time:int=14400,
        test_size:int=2880, val_size:int=2880, horizons:Tuple[int]=(96,
        192, 336, 720))

The ETTh1 dataset monitors an electricity transformer from a region of a province of China including oil temperature and variants of load (such as high useful load and high useless load) from July 2016 to July 2018 at an hourly frequency.

source

​
LongHorizon2

 LongHorizon2 (source_url:str='https://www.dropbox.com/s/rlc1qmprpvuqrsv/a
               ll_six_datasets.zip?dl=1')

This Long-Horizon datasets wrapper class, provides with utility to download and wrangle the following datasets:
ETT, ECL, Exchange, Traffic, ILI and Weather.

  • Each set is normalized with the train data mean and standard deviation.
  • Datasets are partitioned into train, validation and test splits.
  • For all datasets: 70%, 10%, and 20% of observations are train, validation, test, except ETT that uses 20% validation.
Long-Horizon DatasetsM3 dataset