CODEWITHSUNDEEP
rtidyverseone-hot-encoding
Dmytro Fedoriuk
Dmytro Fedoriuk

Reputation: 351

Create dummy variables from all categorical variables in a dataframe

I need to one-encode all categorical columns in a dataframe. I found something like this:

one_hot <- function(df, key) {
  key_col <- dplyr::select_var(names(df), !! rlang::enquo(key))
  df <- df %>% mutate(.value = 1, .id = seq(n()))
  df <- df %>% tidyr::spread_(key_col, ".value", fill = 0, sep = "_") %>% 
  select(-.id)
}

but I can't figure out how to apply it for all categorical columns.

keys <- select_if(data, is.character)[-c(1:2)]
tmp <- map(keys, function(names) reduce(data, ~one_hot(.x, keys)))

throws next error

Error: var must evaluate to a single number or a column name, not a list

upd:

customers <- data.frame(
  id=c(10, 20, 30, 40, 50),
  gender=c('male', 'female', 'female', 'male', 'female'),
  mood=c('happy', 'sad', 'happy', 'sad','happy'),
  outcome=c(1, 1, 0, 0, 0))
customers

after encoding

  id gender.female gender.male mood.happy mood.sad outcome
1 10             0           1          1        0       1
2 20             1           0          0        1       1
3 30             1           0          1        0       0
4 40             0           1          0        1       0
5 50             1           0          1        0       0

Upvotes: 5

Views: 6447

Answers (4)

Oriol Prat
Oriol Prat

Reputation: 1047

Also one-liner with fastDummies package.

fastDummies::dummy_cols(customers)

  id gender  mood outcome gender_male gender_female mood_happy mood_sad
1 10   male happy       1           1             0          1        0
2 20 female   sad       1           0             1          0        1
3 30 female happy       0           0             1          1        0
4 40   male   sad       0           1             0          0        1
5 50 female happy       0           0             1          1        0

Upvotes: 5

Roman
Roman

Reputation: 4989

One-liner with mltools and data.table:

one_hot(as.data.table(customers))

   id gender_female gender_male mood_happy mood_sad outcome
1: 10             0           1          1        0       1
2: 20             1           0          0        1       1
3: 30             1           0          1        0       0
4: 40             0           1          0        1       0
5: 50             1           0          1        0       0

It one-hots all factor variables and has some nice features built in on how to handle NA`s and unused factor levels.

Upvotes: 0

bschneidr
bschneidr

Reputation: 6278

Here's an approach using the recipes package.

library(dplyr)
library(recipes)

# Declares which variables are the predictors
recipe(formula = outcome ~ .,
       data = customers) %>% 
# Declare that one-hot encoding will be applied to all nominal variables
step_dummy(all_nominal(),
           one_hot = TRUE) %>% 
# Based on the previous declarations, apply transformations to the data
# and return the resulting data frame
prep() %>% 
juice()

Upvotes: 2

Jordo82
Jordo82

Reputation: 816

Using the dummies package:

library(dummies)
dummy.data.frame(customers)

  id genderfemale gendermale moodhappy moodsad outcome
1 10            0          1         1       0       1
2 20            1          0         0       1       1
3 30            1          0         1       0       0
4 40            0          1         0       1       0
5 50            1          0         1       0       0

Upvotes: 4

Related Questions