Datasets

The datasets module contains:

• dataset objects, which are in-memory representations of time series datasets. The programmatic way to represent and access datasets in sktime.

• loaders which fetch datasets from data repositories on the internet, and retrieve them as in-memory datasets in sktime compatible formats

• loaders which fetch an individual dataset, usually for illustration purposes

• toy data generators for didactic and illustrative purposes

• utilities to write to, and load from, time series specific file formats

Forecasting datasets

Dataset repositories

Interfaces to dataset repositories, instances of the classes represent different datasets. Downloaded from the sktime huggingface space, cached on first use.

monash.ForecastingData(name)

Monash Forecasting Repository dataset loader.

Individual datasets - onboard

Individual datasets distributed with sktime, available without internet access.

airline.Airline()	Load the airline univariate time series dataset [1].
monash.ForecastingData(name)	Monash Forecasting Repository dataset loader.
hierarchical_sales_toydata.HierarchicalSalesToydata()	Return hierarchical sales toy data to demonstrate hierarchical forecasting.
longley.Longley([y_name])	Load the Longley dataset for forecasting with exogenous variables.
lynx.Lynx()	Load the Lynx dataset for univariate time series forecasting.
m5_competition.M5Dataset([extract_path])	Fetch M5 dataset from https://zenodo.org/records/12636070 .
macroeconomic.Macroeconomic()	Load the US Macroeconomic dataset for multivariate time series forecasting.
pbs.PBS()	Load the Pharmaceutical Benefit Scheme (PBS) univariate time series dataset [Rb1f075efaefb-1].
shampoo_sales.ShampooSales()	Load the Shampoo Sales dataset for univariate time series forecasting.
uschange.USChange([y_name])	Load USChange dataset for forecasting growth rates of consumption and income.

Individual datasets - downloaded

Individual datasets downloadable from the sktime huggingface space, cached on first use.

m5_competition.M5Dataset([extract_path])	Fetch M5 dataset from https://zenodo.org/records/12636070 .
solar.Solar([start, end, normalise, ...])	Load the GB National Solar Estimates dataset for time series forecasting.

Classification datasets

Dataset repositories

Interfaces to dataset repositories, instances of the classes represent different datasets. Downloaded from the sktime huggingface space, cached on first use.

ucr_uea_archive.UCRUEADataset(name[, ...])

UCR-UEA time series classification dataset loader.

Individual datasets - onboard

Individual datasets distributed with sktime, available without internet access.

acsf1.ACSF1([return_mtype])	ACSF1 time series classification problem.
arrow_head.ArrowHead([return_mtype])	ArrowHead time series classification problem.
basic_motions.BasicMotions([return_mtype])	BasicMotions time series classification problem.
gunpoint.GunPoint([return_mtype])	GunPoint time series classification problem.
italy_power_demand.ItalyPowerDemand([...])	ItalyPowerDemand time series classification problem.
japanese_vowels.JapaneseVowels([return_mtype])	JapaneseVowels time series classification problem.
osuleaf.OSULeaf([return_mtype])	OSULeaf time series classification problem.
plaid.PLAID([return_mtype])	PLAID time series classification problem.

Regression datasets

tecator.Tecator()

Load the Tecator time series regression problem.

Dataset loader functions

Loaders are raw functions which return datasets in sktime compatible formats.

For programmatic access to datasets, the dataset objects above should be preferred.

Loaders from dataset repositories

These loaders access dataset repositories on the internet and fetch one or multiple datasets from there, individual datasets specifiable as strings.

These loaders can be used to access reference datasets for benchmarking.

Functions to load and write datasets.

load_forecastingdata(name[, ...])	Fetch forecasting datasets from Monash Time Series Forecasting Archive.
load_fpp3(dataset[, temp_folder])	Load a dataset from the fpp3 package.
load_m5([extract_path, include_events, ...])	Fetch M5 dataset from https://zenodo.org/records/12636070 .
load_UCR_UEA_dataset(name[, split, ...])	Load dataset from UCR UEA time series archive.

Individual datasets

These loaders fetch a commonly used individual dataset, usually for illustration purposes.

Single time series

Functions to load and write datasets.

load_airline()	Load the airline univariate time series dataset [1].
load_longley([y_name])	Load the Longley dataset for forecasting with exogenous variables.
load_lynx()	Load the lynx univariate time series dataset for forecasting.
load_macroeconomic()	Load the US Macroeconomic Data [R7ccb796becb7-1].
load_shampoo_sales()	Load the shampoo sales univariate time series dataset for forecasting.
load_solar([start, end, normalise, ...])	Get national solar estimates for GB from Sheffield Solar PV_Live API.
load_uschange([y_name])	Load MTS dataset for forecasting Growth rates of personal consumption and income.

Panels of time series

Functions to load and write datasets.

load_acsf1([split, return_X_y, return_type])	Load dataset on power consumption of typical appliances.
load_arrow_head([split, return_X_y, return_type])	Load the ArrowHead time series classification problem and returns X and y.
load_basic_motions([split, return_X_y, ...])	Load the BasicMotions time series classification problem and returns X and y.
load_gunpoint([split, return_X_y, return_type])	Load the GunPoint time series classification problem and returns X and y.
load_italy_power_demand([split, return_X_y, ...])	Load ItalyPowerDemand time series classification problem.
load_japanese_vowels([split, return_X_y, ...])	Load the JapaneseVowels time series classification problem.
load_macroeconomic()	Load the US Macroeconomic Data [R7ccb796becb7-1].
load_osuleaf([split, return_X_y, return_type])	Load the OSULeaf time series classification problem and returns X and y.
load_tecator([split, return_X_y, ...])	Load the Tecator time series regression problem and returns X and y.

Toy data generators

Hierarchical time series data

Functions to load and write datasets.

load_hierarchical_sales_toydata()

Return hierarchical sales toy data to demonstrate hierarchical forecasting.

File format loaders and writers

These utilities load and write from time series specific data file formats.

Note: for loading/writing from formats not specific to time series, use common utilities such as pandas.read_csv

Functions to load and write datasets.

load_from_arff_to_dataframe(...[, ...])	Load data from a .arff file into a Pandas DataFrame.
load_from_tsfile(full_file_path_and_name[, ...])	Load time series .ts file into X and (optionally) y.
load_from_tsfile_to_dataframe(...[, ...])	Load data from a .ts file into a Pandas DataFrame.
load_from_ucr_tsv_to_dataframe(...[, ...])	Load data from a .tsv file into a Pandas DataFrame.
load_from_long_to_dataframe(...[, separator])	Load data from a long format file into a Pandas DataFrame.
load_tsf_to_dataframe(full_file_path_and_name)	Convert the contents in a .tsf file into a dataframe.
write_panel_to_tsfile(data, path[, target, ...])	Write an sktime multi-instance dataset to text file in .ts format.
write_dataframe_to_tsfile(data, path[, ...])	Output a dataset in dataframe format to .ts file.
write_ndarray_to_tsfile(data, path[, ...])	Output a dataset in ndarray format to .ts file.
write_tabular_transformation_to_arff(data, ...)	Transform dataset using a tabular transformer and write the result to arff file.
write_results_to_uea_format(estimator_name, ...)	Write the predictions for an experiment in the standard format used by sktime.