Automunge

Latest version: v8.33

Safety actively analyzes 715032 Python packages for vulnerabilities to keep your Python projects secure.

Scan your dependencies

Page 73 of 99

3.73

- performed a walkthrough of various uses of the process_dict entry for 'labelctgy'
- found there was a redundant step performed in shuffleaccuracy function
- removed the redundancy, also cleaned up a few code comment relics

3.72

- found and fixed small bug in feature importance evaluation associated with numeric label sets.
- added convention that label processing under automation does not consider the powertransform parameter.
- (user can still apply powertransform option by passing label column to ptfm in assigncat)

3.71

- As a relic from early implementations, there were a few cases where NArw category columns were given special treatment. This update replaces a few of these edge cases to make more generalizable treatment of NArw columns as consistent with other categories.
- 1) updated methods for convention that label sets are returned without NArw columns. Previously we derived the NArw column for labels if NArw_marker was activated and then deleted. New way is to have dedicated root category family trees for labels under automation which don't have NArw in their trees. This way a user still has option to derive an NArw column for label sets if desired by redefining the label trees in the transformdict parameter.
- 2) in the population of data structures for postprocess_dict['origcolumn'], we had given special treatment to using NArw column as an entry to ['columnkey']. This no longer required, now NArw columns treated just like any other.
- 3) in feature importance dimensionality reduction, we previously had given special treatment to NArw columns, this convention is now scrubbed.
- also corrected process_dict entry for ord3, ordd
- also removed an uneccesary logic test in postmunge infill application associated with infilliterate
- also removed a few unneccesary logic tests in automunge and postmunge infill application
- Updated convention for various infill options to add support for categories with MLinfilltype 'exclude', so although 'exclude' not included in MLinfill, it is elligible for other infill types. The only MLinfilltype exlcuded from all infill is 'boolexclude', the one used for NArw.

3.70

- some cleanup of infill methods
- moved some steps associated with MLinfill parameter into a support function for clarity
- renamed an infill item for consistency
- replaced an infill methods test for 'NArw' category to a test of MLinfilltype for better generality
- removed an unused 'if' indentation in postmunge(.)
- removed an unused function (superseded)
- cleaned up code comments in several places

3.69

- fixed edge case unintentional overwrite for data object returned in postprocess_dict['postprocess_assigninfill_dict']
- now able to use this object to replace a postmunge function call for (slightly) more efficent infill operation
- scrubbed reference to the superseded 'label' MLinfilltype
- removed an unused parameter from the assembletransformdict function to avoid temptation of repurposing

3.68

- added ML_cmnd support for 'n_jobs' parameter for scikit Random Forest and PCA training
- (n_jobs allows user to parallelize training accross concurent processor threads)
- added ML_cmnd support for 'criterion' parameter for scikit Random Forest training
- (the exclusion of this parameter turned out to have been result of a silly mixup)
- added new section for concise sorted list of root categories to READ ME

Page 73 of 99

Links

Releases

Has known vulnerabilities

© 2025 Safety CLI Cybersecurity Inc. All Rights Reserved.