datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
exchange_rate
Exchange Rate Dataset
Part of the TS Arena benchmarking suite for time-series foundation models.
Description
Daily exchange rates of 8 countries (1990-2016)
Statistics
Total samples: 7588
Features: 8
Train samples: 5311
Validation samples: 759
Test samples: 1518
Frequency: 1 day
Preprocessing
Standard normalization (zero mean, unit variance)
Scaler fitted on training data only
Train/Val/Test split: 70%/10%/20%
Usage
from datasets… See the full description on the dataset page: https://huggingface.co/datasets/ts-arena/exchange_rate.weather
Weather Dataset
Part of the TS Arena benchmarking suite for time-series foundation models.
Description
Max Planck Institute Weather Station - 2020
Statistics
Total samples: 52696
Features: 21
Train samples: 36887
Validation samples: 5269
Test samples: 10540
Frequency: 10 minutes
Preprocessing
Standard normalization (zero mean, unit variance)
Scaler fitted on training data only
Train/Val/Test split: 70%/10%/20%
Usage
from datasets… See the full description on the dataset page: https://huggingface.co/datasets/ts-arena/weather.traffic
Traffic Dataset
Part of the TS Arena benchmarking suite for time-series foundation models.
Description
Road occupancy rates from 862 sensors in San Francisco Bay Area
Statistics
Total samples: 17544
Features: 862
Train samples: 12280
Validation samples: 1755
Test samples: 3509
Frequency: 1 hour
Preprocessing
Standard normalization (zero mean, unit variance)
Scaler fitted on training data only
Train/Val/Test split: 70%/10%/20%
Usage… See the full description on the dataset page: https://huggingface.co/datasets/ts-arena/traffic.illness
Illness Dataset
Part of the TS Arena benchmarking suite for time-series foundation models.
Description
Influenza-like Illness weekly from CDC
Statistics
Total samples: 966
Features: 7
Train samples: 676
Validation samples: 96
Test samples: 194
Frequency: 1 week
Preprocessing
Standard normalization (zero mean, unit variance)
Scaler fitted on training data only
Train/Val/Test split: 70%/10%/20%
Usage
from datasets import load_dataset
#… See the full description on the dataset page: https://huggingface.co/datasets/ts-arena/illness.
