recpack.scenarios.splitters.StrongGeneralizationSplitter

class recpack.scenarios.splitters.StrongGeneralizationSplitter(in_frac=0.7, seed=None, error_margin=0.01)

Randomly splits the users into two sets so that interactions for a user will always occur only in one split.

Parameters

in_frac (float, optional) – The fraction of interactions that are assigned to the first value in the output tuple. Defaults to 0.7.
seed (int, optional) – Seed the random generator. Set this value if you require reproducible results.
error_margin (float, optional) – The allowed error between in_frac and the actual split fraction. Defaults to 0.01.

Methods

Randomly splits the users into two sets so that

Attributes

`identifier`	String identifier of the splitter object, contains name and parameter values.
`name`	The name of the splitter.

property identifier: String identifier of the splitter object, contains name and parameter values.

split(data)

Randomly splits the users into two sets so that: interactions for a user will always occur only in one split.

Parameters: data (InteractionMatrix) – Interaction matrix to be split.
Returns: A 2-tuple containing the data_in and data_out matrices.
Return type: Tuple[InteractionMatrix, InteractionMatrix]