Ability to pass through feature(s) to prediction output without use in pipeline
- Linguagem predominante
- Python
- Estrelas
- 850
- Forks
- 96
- Métricas de merge de PRs
- Nenhum PR com merge em 30d
Descrição
**Description**
Support the ability to designate one or more features as "pass through" features, meaning they will _not_ be used by any transformers or by the estimator, but will be provided as additional column(s) alongside predictions in the pipeline output.
**Use cases**
* ID column: we want an ID column to appear alongside predictions for convenience, but we don't want to make that the dataframe index. The ID column should not be provided to the estimator as a feature.
* Custom objective: as in our lead scoring demo, make "cost" column available to objective function, but don't provide that column to the estimator as a feature.
**Notes**
Perhaps this should be implemented using a woodwork `pass_through_feature` tag.
Tagging @rpeck since he brought this up.
Guia de contribuição
Avaliação
Esta issue ainda não foi avaliada.