CODEWITHSUNDEEP
r
Manu Vats
Manu Vats

Reputation: 147

Rename categorical levels to reduce the number of levels

I have a categorical column which has around 1200 levels in a data-set of around 78000 records. I want to reduce the levels on the basis of occurrences. For eg. -:

And so on.

I don't want to group less frequently occurring levels into 'Others' as it would hide a lot of important levels.
Following is the dataframe for example.

df=data.frame(
  ID = c(1:10),
  Name = c("Jack", "Mike","Jack", "Mike","Jack", "Mike", "Tom", "Tom", "Smith", "Tony")
)

Here I would like to reduce the levels of column 'Name' by-:

Can anyone help me to do it in R?

Upvotes: 3

Views: 306

Answers (3)

akrun
akrun

Reputation: 887951

We can use fcase from the devel version of data.table (1.12.9), which would also do evaluation lazily

library(data.table)
setDT(df)[, NewName := fcase(.N >=3, 'A',
                             .N >=2 & .N < 3, 'B',
                             default = 'C'), Name][]
#    ID  Name NewName
# 1:  1  Jack       A
# 2:  2  Mike       A
# 3:  3  Jack       A
# 4:  4  Mike       A
# 5:  5  Jack       A
# 6:  6  Mike       A
# 7:  7   Tom       B
# 8:  8   Tom       B
# 9:  9 Smith       C
#10: 10  Tony       C

Or using base R with findInterval

with(df, rev(LETTERS[1:3])[findInterval(table(Name)[Name], 2:3) + 1])
#[1] "A" "A" "A" "A" "A" "A" "B" "B" "C" "C"

Upvotes: 0

G5W
G5W

Reputation: 37661

A base R solution using table

NameCount = table(df$Name)[df$Name]
NewName = rep("C", length(NameCount))
NewName[NameCount >= 2] = "B"
NewName[NameCount >= 3] = "A"
NewName
 [1] "A" "A" "A" "A" "A" "A" "B" "B" "C" "C"

Upvotes: 2

Ronak Shah
Ronak Shah

Reputation: 389325

We can use add_count to count instances of Name and then check conditions with case_when

library(dplyr)

df %>% 
  add_count(Name) %>%
  mutate(NewName = case_when(n >= 3 ~'A', 
                             n >= 2 & n < 3 ~'B', 
                             TRUE ~'C')) %>%
  select(-n, -Name)

#     ID NewName
#   <int> <chr>  
# 1     1 A      
# 2     2 A      
# 3     3 A      
# 4     4 A      
# 5     5 A      
# 6     6 A      
# 7     7 B      
# 8     8 B      
# 9     9 C      
#10    10 C

Upvotes: 3

Related Questions