- incorporated feature importance evaluation for postmunge, available by passing featureeval=True
2.40
- Added support for passed numpy arrays. Input datasets can now be either pandas or numpy objects.
2.39
- a few more printouts for feature importance evaluation for clarity - update method for calculating bins for powers of 10 - added two new entries to postprocess_dict listing returned columns for labels and ID sets for cases where user elected return of numpy arrays and later needs to identify column. You know, logistics.
2.38
- user passed trasnformdict without a replacement primitive now assigned a passthrough category (excl) to auntsuncles - revisited a few of the logic steps performed in evalcategory - removed a few steps of validation comparing train and test data column configurations for simplicity and run time. Method now makes assumption without verification that test data is consistently formatted as train data. - cleanup of a few variable naming conventions
2.37
fixed bug introduced in 2.36 associated with printouts for postmunge
2.36
- Frequency levelizer option added to postmunge. - fixed nbr3 definition - Printout returned label set before Automunge complete - assign powertransform to specific columns in assigncat via 'eval' - validate passed transformdict to determine if a root defined without a replacement column - validate passed ML_cmnd and fixed format if missing entries