Skip to content

Latest commit

 

History

History
36 lines (27 loc) · 1.83 KB

CHANGELOG.md

File metadata and controls

36 lines (27 loc) · 1.83 KB
Raw

Change Log

All notable changes to this project will be documented in this file. This project adheres to Semantic Versioning.

[dev] (unreleased)

  • ...

[1.2.0]

Improvements:

  • Handle pd.Series that are IntegerArrays e.g. dtype Int64

[1.1.1]

Improvements:

  • When installing ppscore automatically install dependencies. (#29)

[1.1.0]

Default changes:

  • In pps.score changed the default of random_seed to 123 in order to enable reproducible results when comparing scores that are generated via pps.score or pps.matrix. The old behavior of drawing a random_seed per default was confusing. (#32)

Improvements:

  • Improve error message when using the wrong API (#31)
  • Automatic error catching for convenience with optional error inspection/debugging: added catch_errors to pps.score and added corresponding case unknown_error

[1.0.0]

Breaking changes:

  • The case of the calculation for regression or classification only depends on the data types and not the cardinality of a column any more. Also, it is not possible any more to pass the task to pps.score. This removes confusion about the inference of the final task. Some other special cases like feature_is_id still review the cardinality of a column.
  • The return format of pps.matrix changed to a tidy dataframe
  • Changed the format of the PPS dict. It now includes information about errors and special cases in the case field. Also, there is is_valid_score

Improvements:

  • Added new kwargs to pps.score e.g. cross_validation, random_seed and invalid_score
  • Added error messages when the input arguments are invalid
  • Added more tests
  • Added CHANGELOG