Skip to content

Improve spatial splitting functionality #546

Description

@RaczeQ
  • Add option to pass GeoSeries with points or list of Shapely points as the input_gdf parameter and allow for target_column to be a list of values (or numpy array / pandas Series) of the same length as input.
  • Add option to use custom geometries instead of only H3 cells. Leave H3 cells logic as default, and add new function where user can pass GeoSeries with geometries to use as grouping "cells". Raise an error if points are outside provided geometries.
  • Refactor spatial_split_points function to split into N splits, not only train/test/validation. Add parameter for passing a list of ratios and split into the same number of splits.
  • Combine previous refactor with option to create K-Folds (eg generate 5 splits and return 5 folds with 1:4 ratio)

Metadata

Metadata

Assignees

Labels

enhancementNew feature or request

Fields

No fields configured for Feature.

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions