step_mutate or step_mutate_at using custom function

Machine Learning and Modeling
1

Hi there. I am trying to create a recipe with a step function that takes the input character variable and bins it into various categories depending on the value of the variable. For example

my_custom_function(word){
     if(str_detect(word, "a")) return("A")
     if(str_detect(word, "b")) return("B")
     return("C")

main_rec <- recipe(y ~ word,  df) |>
      step_mutate_at(word, fn = list(my_ custom_function))

# OR

main_rec <- recipe(y ~ word,  df) |>
      step_mutate(word =  as.factor(map_chr(word, .f = my_custom_function)))

It throws an error either not recognizing the custom function ot map_chr or str_detect.

What am I doing wrong?

Any help is appreciated.

2

Using a mapping function inside of step_mutate() might be pushing the boundaries of what it can do (esp in parallel).

You could change your function to a case_when() and achieve the same thing. There is also step_regex(); depending on what you do with the return values, that might solve it for you.