`simpleml.pipelines.projected_splits`

Module for dataset projection into pipelines. Defines transfer objects returned from pipelines

Module Contents

`IdentityProjectedDatasetSplit`	Straight passthrough variety of projection (ie projected split == dataset split)
`IndexBasedProjectedDatasetSplit`	Index based subset. Compatible with dataset splits that support indexing
`ProjectedDatasetSplit`	Transfer object to pass dataset splits through pipelines

class simpleml.pipelines.projected_splits.IdentityProjectedDatasetSplit(dataset, split)[source]

Straight passthrough variety of projection (ie projected split == dataset split)

Parameters

apply_projection(self, dataset_split)[source]

Identity return

class simpleml.pipelines.projected_splits.IndexBasedProjectedDatasetSplit(indices, **kwargs)[source]

Index based subset. Compatible with dataset splits that support indexing

apply_projection(self, dataset_split)[source]

Index subset return

classmethod indexing_method(cls, df, *args, **kwargs)[source]: Infer indexing method to use based on type

class simpleml.pipelines.projected_splits.ProjectedDatasetSplit(dataset, split)[source]

Transfer object to pass dataset splits through pipelines

Contains a reference to the dataset and internal logic to project the split (references the dataset on each call to avoid mutability issues)

Wraps the normal Split object but delegates behavior so can be used interchangeably

Parameters

__getattr__(self, attr)[source]: Passthrough to treat a projected split like a normal split

abstract apply_projection(self, dataset_split)[source]

Main method to apply projection logic on the dataset split Returns a new Split with the data subset

property dataset_split(self)[source]

Passthrough method to retrieve the raw split

property projected_split(self)[source]

Wrapper property to retrieve the dataset split and manipulate into a projected split. Returns a split object already parsed