Datasets

Functions to obtain datasets.

Fetching real datasets

The following functions are used to retrieve specific functional datasets:

skfda.datasets.fetch_aemet()	Load the Spanish Weather dataset.
skfda.datasets.fetch_gait()	Load the GAIT dataset.
skfda.datasets.fetch_growth()	Load the Berkeley Growth Study dataset.
skfda.datasets.fetch_handwriting()	Load the HANDWRIT dataset.
skfda.datasets.fetch_mco()	Load the mithochondiral calcium overload (MCO) dataset.
skfda.datasets.fetch_medflies()	Load the Medflies dataset.
skfda.datasets.fetch_nox()	Load the NOx dataset.
skfda.datasets.fetch_octane()	Load near infrared spectra of gasoline samples.
skfda.datasets.fetch_phoneme()	Load the phoneme dataset.
skfda.datasets.fetch_tecator()	Load the Tecator dataset.
skfda.datasets.fetch_weather()	Load the Canadian Weather dataset.

Those functions return a dictionary with at least a “data” field containing the instance data, and a “target” field containing the class labels or regression values, if any.

In addition datasets can be downloaded from CRAN and the UCR:

skfda.datasets.fetch_cran(name, package_name, *)	Fetch a dataset from CRAN.
skfda.datasets.fetch_ucr()	Fetch a dataset from the UCR/UEA repository.

Datasets from CRAN are not in a standardized format. Datasets from the UCR are in the same format as the specific datasets, but often have an explicit test set, accessible as “data_test” and “target_test”.

Making synthetic datasets

The following functions are used to make synthetic functional datasets:

skfda.datasets.make_gaussian([n_samples, ...])	Generate Gaussian random fields.
skfda.datasets.make_gaussian_process([...])	Generate Gaussian process trajectories.
skfda.datasets.make_sinusoidal_process([...])	Generate sinusoidal proccess.
skfda.datasets.make_multimodal_samples([...])	Generate multimodal samples.
skfda.datasets.make_multimodal_landmarks([...])	Generate landmarks points.
skfda.datasets.make_random_warping([...])	Generate random warping functions.