Target mean encoder implementation in R

target_mean_encoder.R

library(dplyr)

# creating a toy dataset
data = tibble(vehicle = c("car", "bus", "bike", "bus", "car", "bike"),
              target = c(23,34,56,78,33,65))
                  
# print dataframe
data

# OUTPUT
# A tibble: 6 x 2
#  vehicle target
#  <chr>     <dbl>
#1 car          23
#2 bus          34
#3 bike         56
#4 bus          78
#5 car          33
#6 bike         65

# Creating the lookup vector for the variable to be encoded
target_mean_enc_lookup <- data %>% 
    group_by(vehicle) %>% 
    summarise(mean_target = mean(target)) %>% 
    ungroup() %>% 
    arrange(mean_target) %>% 
    pull(vehicle) %>% 
    setNames(1:length(.), .)

# printing the lookup vector
target_mean_enc_lookup

# OUTPUT
# car bus bike
#  1   2   3

# Encoding the vehicle variable
data %>% 
    mutate(
        vehicle_target_mean_enc = unname(target_mean_enc_lookup[data$vehicle])
    )

# print dataframe
data

# OUTPUT
# A tibble: 6 x 4
#  vehicle    target      vehicle_target_mean_enc   
#  <chr>       <dbl>               <int>
#1 car          23                   1
#2 bus          34                   2
#3 bike         56                   3
#4 bus          78                   2
#5 car          33                   1
#6 bike         65                   3