Write basic feature stats to tables after feature generation
- Dominant language
- Jupyter Notebook
- Stars
- 201
- Forks
- 62
- PR merge metrics
- No merged PRs in 30d
Description
@saleiro @ecsalomon and I went on a bit of a dive this morning following my exploration of the https://github.com/tdda/tdda repository . The team has talked about "premodeling" before, that could expand in scope to be as big as postmodeling. Right now, however, I think we should figure out whats the simplest possible thing that could be useful as a beachhead for this premodeling. And in our discussion I think we have something.
After feature generation (both pre-imputation and post-imputation), create feature stats in table/s that encompass basic min/max/nulls stats for the created features. These would always reflect the current state of the entire feature table (e.g. we recompute each time), provided that this doesn't take too long.
Contributor guide
Assessment
This issue has not been assessed yet.