1

有没有办法在创建虚拟变量时指定参考级别step_dummy()？我可以通过设置one_hot = TRUE然后删除参考列来做到这一点，但想知道是否可以在其内部step_dummy()指定

library(tidymodels)
#> Registered S3 method overwritten by 'tune':
#>   method                   from   
#>   required_pkgs.model_spec parsnip

data(okc)

# level "anything" is the reference level
recipe(Class ~ ., data = okc) %>%
  step_dummy(diet) %>%
  prep() %>%
  bake(new_data = NULL) %>%
  select(starts_with("diet")) %>%
  names()
#> Warning: There are new levels in a factor: NA
#>  [1] "diet_halal"               "diet_kosher"             
#>  [3] "diet_mostly.anything"     "diet_mostly.halal"       
#>  [5] "diet_mostly.kosher"       "diet_mostly.other"       
#>  [7] "diet_mostly.vegan"        "diet_mostly.vegetarian"  
#>  [9] "diet_other"               "diet_strictly.anything"  
#> [11] "diet_strictly.halal"      "diet_strictly.kosher"    
#> [13] "diet_strictly.other"      "diet_strictly.vegan"     
#> [15] "diet_strictly.vegetarian" "diet_vegan"              
#> [17] "diet_vegetarian"

# all 18 diet levels included
recipe(Class ~ ., data = okc) %>%
  step_dummy(diet, one_hot = TRUE) %>%
  prep() %>%
  bake(new_data = NULL) %>%
  select(starts_with("diet")) %>%
  names()
#> Warning: There are new levels in a factor: NA
#>  [1] "diet_anything"            "diet_halal"              
#>  [3] "diet_kosher"              "diet_mostly.anything"    
#>  [5] "diet_mostly.halal"        "diet_mostly.kosher"      
#>  [7] "diet_mostly.other"        "diet_mostly.vegan"       
#>  [9] "diet_mostly.vegetarian"   "diet_other"              
#> [11] "diet_strictly.anything"   "diet_strictly.halal"     
#> [13] "diet_strictly.kosher"     "diet_strictly.other"     
#> [15] "diet_strictly.vegan"      "diet_strictly.vegetarian"
#> [17] "diet_vegan"               "diet_vegetarian"

# force diet_vegan to be reference level
recipe(Class ~ ., data = okc) %>%
  step_dummy(diet, one_hot = TRUE) %>%
  step_select(-diet_vegan) %>%
  prep() %>%
  bake(new_data = NULL) %>%
  select(starts_with("diet")) %>%
  names()
#> Warning: There are new levels in a factor: NA
#>  [1] "diet_anything"            "diet_halal"              
#>  [3] "diet_kosher"              "diet_mostly.anything"    
#>  [5] "diet_mostly.halal"        "diet_mostly.kosher"      
#>  [7] "diet_mostly.other"        "diet_mostly.vegan"       
#>  [9] "diet_mostly.vegetarian"   "diet_other"              
#> [11] "diet_strictly.anything"   "diet_strictly.halal"     
#> [13] "diet_strictly.kosher"     "diet_strictly.other"     
#> [15] "diet_strictly.vegan"      "diet_strictly.vegetarian"
#> [17] "diet_vegetarian"

由reprex 包于 2021-11-19 创建(v2.0.1)

4

1 回答 1

1

来自step_dummy()文档

默认情况下，排除的虚拟变量（即参考单元格）将对应于被转换的无序因子的第一级。

我们可以使用step_relevel()ref_level通过设置参数来创建新的参考级别。

library(tidymodels)

data(okc)

recipe(Class ~ ., data = okc) %>%
  step_relevel(diet, ref_level = "vegan") %>%
  step_dummy(diet) %>%
  prep() %>%
  bake(new_data = NULL) %>%
  select(starts_with("diet")) %>%
  names()
#> Warning: There are new levels in a factor: NA
#>  [1] "diet_anything"            "diet_halal"              
#>  [3] "diet_kosher"              "diet_mostly.anything"    
#>  [5] "diet_mostly.halal"        "diet_mostly.kosher"      
#>  [7] "diet_mostly.other"        "diet_mostly.vegan"       
#>  [9] "diet_mostly.vegetarian"   "diet_other"              
#> [11] "diet_strictly.anything"   "diet_strictly.halal"     
#> [13] "diet_strictly.kosher"     "diet_strictly.other"     
#> [15] "diet_strictly.vegan"      "diet_strictly.vegetarian"
#> [17] "diet_vegetarian"
于 2021-11-19T19:39:57.427 回答