dummy_to_cols: Fast Encode Non-Numeric Variables into Dummy Columns

In yu45020/encodedummy: Fast Encode Non-Numeric Variables into Dummy Columns

Description

In a data frame, if there are columns in character or factor format, encodedummy_col can convert them into dummy columns, and each columns has either 1 or 0. Require data.table.

Usage

dummy_to_cols(data, drop_first_level=FALSE,
keep_origin_cols=FALSE, sep_char='=', inplace=FALSE)

Arguments

data	Either a data.frame or data.table
drop_first_level	Whether drop the first level of category
keep_origin_cols	Whether keep the input data's non-numeric columns
sep_char	Character string between column names and unique category/level. Example: sep_char="=", and a column named 'Col A' with levels 'A','B'. Then the output columns will be named as 'Col A=A', 'Col A=B'.
inplace	Whether modify the data directly. Input data will be converted into data.table, so FALSE means a new copy is created and added new dummy columns, while the input data is not changed. TRUE means the input data is modified directly in memory, so no new object will be created. Details can be found by ?data.table::copy

Value

Returns a data.table object.

Examples

library(data.table)
get_test_data=function(N=1e5,K=6){
   set.seed(1)
   DT <- data.table(
     id1 = factor(sample(letters[1:K], N, TRUE)),
     id3 = factor(sample(LETTERS[1:K], N, TRUE)),
     id4 = sample(K, N, TRUE),
     id5 = factor(sample(K, N, TRUE)),
     id6 = sample(N/K, N, TRUE),
     v1 =  factor(sample(5, N, TRUE)),
     v3 =  sample(round(runif(100,max=100),4), N, TRUE)
  )
 }

DT = get_test_data(1e5,6)
DT_ = copy(DT)
print(format(object.size(DT), units = 'Mb'))

# number of unique values in each column
print(DT[,lapply(.SD, uniqueN)])
system.time(
  DT_new <- encodedummy_col(DT, drop_first_level=FALSE,
  keep_origin_cols=FALSE, sep_char='=',inplace=FALSE)
)
# inplace is FALSE, so the input data is not modified.
# Instead, a new object is created.
identical(DT,DT_) # TRUE
identical(DT,DT_new)  # FALSE

yu45020/encodedummy documentation built on May 23, 2019, 1:25 p.m.