gramex
gramex copied to clipboard
Explicit, friendly keys names for MLhandler configuration
Should we consider explicit, friendly names in configuration?
- target_column for target_col
- exclude_columns for exclude
- categorical_columns for cats
cc @jaidevd @sanand0
Consistency is good. *_column
as a convention is good.
This is early stages, so it's OK to make backward incompatible changes.
If we adopt some standard (e.g. sklearn), that would be even better. Is there a standard we could adopt @jaidevd / @pratapvardhan?
scikit-learn tends to use X for input columns, y for target. Although, mostly usage is with positional argument and not named. So, X, y is not so widely used aswell.
To exclude or numeric columns in scikit we usually use pipeline steps, with ColumnTransformer, make_column_selector etc.
So not sure, it there's something from scikit that can be used at root configuration.