Keyword arg for predicted column name

[janosh]

The default name target + " predicted" for the new column appended to a dataframe by mat_pipe.predict() contains a space and prevents the use of dot notation to access the column. Would you accept a PR to make this suffix configurable or maybe even change the default to target + _predicted" or target + _pred"?

[ardunn]

Hey @janosh I'd accept a PR to make it configurable!

I think this should take form in two parts:
(1) setting the default in the DFMLAdaptor base class somehow (preferred), or manually adding it to both adaptors args (not preferred). Anyway it should just be an init arg to each adaptor

(2) a powerup kwarg to the get_preset_config for MatPipe.from_preset method.

[janosh]

Are you sure this should be part of get_preset_config()? Seems too unimportant to make it part of a preset.

[ardunn]

Yes, in the sense it should be a powerup (kwarg), not a preset itself. The "powerups" are intended to be common options that are applied to the pre-defined presets. The idea is that if you want to enable some common options while using a preset (e.g., cache the features in some file, set the number of jobs, etc.) you don't need to specify an entire custom pipeline for MatPipe.

For example, n_jobs is a powerup which can be applied to any of the presets to define the number of jobs for featurization and learning across the entire pipeline. Similarly, the target_output_col (or whatever name) could be a powerup which sets the output col name.

So without making it a powerup, if you wanted to set the output col name, you'd need to either define and entire custom pipeline (bad, lots of work [relatively]) or get the preset config and then replace the AutoFeaturizer with a custom one (less bad, but still not intuitive).

Making it a powerup in get_preset_config, you could just:

pipe = MatPipe.from_preset("express", target_output_col="custom column name")

[janosh]

I see what you mean. I'll add that.