lukezaruba / stco

A Python package for assessing the temporal stability of spatial clustering methods via spatial overlay analysis.

Luke Zaruba

Spatial Data Science • Data Engineering • Backend Development

     

stco's Issues

Change how ADBSCAN instance is created/passed into TemporalADBSCAN

Rather than passing in ADBSCAN instance as param into TemporalADBSCAN, use kwargs for getting parameters. Have defaults similar to ADBSCAN's.

Min_Samples param is already overwritten to 1% of each period's total rows, make this param a percentage rather than a raw number.

Allows for instantiation of ADBSCAN for each period rather than copying and changing min samples for each.

Documentation & Examples

• Update README
• More detailed example notebooks
• Documentation (Sphinx? Markdown? Wiki?)

CI/CD

• Build, Publish process to PyPI through GitHub Action
• Dependent on Lint/Format/Test passing
• Pre commit runs Lint/Format/Tests, as well

Default Params for ADBSCAN

Look into how to select defaults for:

1. Eps

• Percentage of Latitude/Y Delta
• Density in conjunction with Min Samples

2. Min Samples

• Percentage of snapshot records
• Density in conjunction with Eps/Radius

From ESDA docs:

$$Density = {Min Samples \over \pi * Eps^2} $$

Just need 2/3 to get the other