`recipes::step_num2factor()` 错误 - tidymodels

Question

在使用 tidymodels 方法定义 recipe() 后调用 recipes::prep() 时出现错误。看来我在配方定义期间误用了 recipes::step_num2factor()，但我不明白出了什么问题。

加载包

library(tidyverse)  # data wrangling
library(tidymodels)  # modelling

提供数据

data <- 
  tibble::tribble(
  ~Survived, ~Pclass,
         1L,      1L,
         1L,      2L,
         0L,      3L
  )

定义配方

titanic_recipe <- 
  
  # define model formula:
  recipe(Survived ~ Pclass, data = data) %>%
  
  # convert numeric outcome to nominal (factor):
  step_num2factor(Survived,
                  levels = c("dead", "alive"))   

准备菜谱

prep(titanic_recipe)  # THROWS ERROR

上面的live抛出这个错误：

Error: Assigned data `map_df(...)` must be compatible with existing data.
x Existing data has 3 rows.
x Assigned data has 2 rows.
ℹ Only vectors of size 1 are recycled.

我不明白为什么会出现这个错误。

Answer 1

因子水平不能为零，它们应该从 1 开始。因此，您可以向 step() 函数添加一个 transform 参数以添加 1。下面的修改工作正常

titanic_recipe <- 
  recipe(Survived ~ Pclass, data = data) %>%
  step_num2factor(Survived,
                  levels = c("dead", "alive"),
                  transform = function(x) x+1)   
titanic_recipe

prep(titanic_recipe) %>% juice()

# A tibble: 3 x 2
  Pclass Survived
   <int> <fct>   
1      1 alive   
2      2 alive   
3      3 dead