R recipes step_num2factor 将数字转换为因数

step_num2factor(
  recipe,
  ...,
  role = NA,
  transform = function(x) x,
  trained = FALSE,
  levels,
  ordered = FALSE,
  skip = FALSE,
  id = rand_id("num2factor")
)

其他虚拟变量和编码步骤：step_bin2factor() , step_count() , step_date() , step_dummy_extract() , step_dummy_multi_choice() , step_dummy() , step_factor2string() , step_holiday() , step_indicate_na() , step_integer() , step_novel() , step_ordinalscore() , step_other() , step_regex() , step_relevel() , step_string2factor() , step_time() , step_unknown() , step_unorder()

library(dplyr)
data(attrition, package = "modeldata")

attrition %>%
  group_by(StockOptionLevel) %>%
  count()
#> # A tibble: 4 × 2
#> # Groups:   StockOptionLevel [4]
#>   StockOptionLevel     n
#>              <int> <int>
#> 1                0   631
#> 2                1   596
#> 3                2   158
#> 4                3    85

amnt <- c("nothin", "meh", "some", "copious")

rec <-
  recipe(Attrition ~ StockOptionLevel, data = attrition) %>%
  step_num2factor(
    StockOptionLevel,
    transform = function(x) x + 1,
    levels = amnt
  )

encoded <- rec %>%
  prep() %>%
  bake(new_data = NULL)

table(encoded$StockOptionLevel, attrition$StockOptionLevel)
#>          
#>             0   1   2   3
#>   nothin  631   0   0   0
#>   meh       0 596   0   0
#>   some      0   0 158   0
#>   copious   0   0   0  85


# an example for binning

binner <- function(x) {
  x <- cut(x, breaks = 1000 * c(0, 5, 10, 20), include.lowest = TRUE)
  # now return the group number
  as.numeric(x)
}

inc <- c("low", "med", "high")

rec <-
  recipe(Attrition ~ MonthlyIncome, data = attrition) %>%
  step_num2factor(
    MonthlyIncome,
    transform = binner,
    levels = inc,
    ordered = TRUE
  ) %>%
  prep()

encoded <- bake(rec, new_data = NULL)

table(encoded$MonthlyIncome, binner(attrition$MonthlyIncome))
#>       
#>          1   2   3
#>   low  749   0   0
#>   med    0 440   0
#>   high   0   0 281

# What happens when a value is out of range?
ceo <- attrition %>%
  slice(1) %>%
  mutate(MonthlyIncome = 10^10)

bake(rec, ceo)
#> # A tibble: 1 × 2
#>   MonthlyIncome Attrition
#>   <ord>         <fct>    
#> 1 NA            Yes