recipes
recipes copied to clipboard
Pipeable steps for feature engineering and data preprocessing to prepare for modeling
I learned about `all.names()` a couple of days ago. And it works wonders in recipes! The previous issues we had with long formulas was that `fun_calls()` was recursive, and it...
``` r library(tidymodels) set.seed(12345) dat % step_interact(~ starts_with("race_"):starts_with("hisp_") + starts_with("race_"):starts_with("cat")) %>% prep() colnames(bake(rec, new_data = dat)) #> [1] "outcome" "race_o" "race_w" #> [4] "hisp_nhisp" "cat_b" "cat_c" #> [7] "cat_d" "cat_e"...
Add an option, perhaps called `.pkgs`, and update the `required_pkgs()` method for this step. See [this post](https://community.rstudio.com/t/adding-packages-to-the-recipes-required-pkgs/181216)/
There's a new package on CRAN called `offsetreg`, which extends `tidymodels` with some additional model types and engines that use an offset, i.e., a variable which should always have a...
## The problem Long formulas (with many terms; I don't _believe_ the length of column names matters) cause `recipe()` to error due to C stack usage. Apologies if this is...
what the title says. I forget too often that I should do this
For documentation (see page below), the argument `keep_original_cols` defaults to `TRUE` not `FALSE`. Of course, I personally prefer `TRUE` as the default behavior. https://recipes.tidymodels.org/reference/step_interact.html
The `...` are used throughout the package for different things. Many of the steps define them themselves where it would probably be better if they were to be inherited. Other...
right now deprecated functions live alongside the other files. we should follow the style guide https://style.tidyverse.org/package-files.html and move those into their own files to avoid developer confusion
A number of steps use a `options` argument, which is supposed to be a named list of argument, that we pass to an underlying function. This should be refactored for...