I want to group_by similar results (not unique) and I don\'t know how to do it.
I mean, I have a df with a column called \'name\' that has similar results like: ARPO, AR
The function below uses agrepl
to get similar strings, given a threshold thresh
. And returns an integer vector of positions where the first in a group was found.
The test data is the data in NoeliaNC's answer.
library(dplyr)
similarGroups <- function(x, thresh = 0.8){
grp <- integer(length(x))
name <- x
for(i in seq_along(name)){
if(!is.na(name[i])){
sim <- agrepl(x[i], x, ignore.case = TRUE, max.distance = 1 - thresh)
k <- which(sim & !is.na(name))
grp[k] <- i
is.na(name) <- k
}
}
grp
}
similarGroups(df[['name']])
# [1] 1 1 1 1 1 6 6 8 8 6 8 8 6 1 1 8 1 6 6 6 1 8 1 8 1
Now apply the function to grouping the dataframe.
df %>%
mutate(group = name[similarGroups(name)]) %>%
count(group)
## A tibble: 3 x 2
# group n
#
#1 ARPO 11
#2 BBBBB 7
#3 JJJJ 7
Edit
Another way is to use the stringsim
function in package stringdist
. It features several distance/similarity measures, that can be tested to see which one gives better results.
similarGroups2 <- function(x, thresh = 0.8, method = "soundex"){
grp <- integer(length(x))
name <- x
x <- tolower(x)
for(i in seq_along(name)){
if(!is.na(name[i])){
sim <- stringdist::stringsim(x[i], x, method = method)
k <- which(sim > thresh & !is.na(name))
grp[k] <- i
is.na(name) <- k
}
}
grp
}
df %>%
mutate(group = name[similarGroups2(name, thresh = 0.7, method = "jw")]) %>%
count(group)
## A tibble: 4 x 2
# group n
#
#1 Antonio Gomez 3
#2 ARPO 11
#3 BBBBB 7
#4 JJJJ 7
df <- tibble::tribble(
~name, ~number, ~ind,
'Antonio Gomez', 1234, 1,
'Antonio Sanches', 5678, 2,
'Antonio Ruiz', 9089, 3,
"ARPO", "405162", 5,
"ARPO S.L.", "504653", 22,
"ARPOS", "900232", 1,
"ARPO", "504694", 12,
"ARPO", "400304", 42,
"JJJJ", "401605", 2,
"JJJJ", "900029", 31,
"BBBBB", "400090", 25,
"BBBBB", "403004", 33,
"JJJJ", "900222", 2,
"BBBBB", "403967", 11,
"BBBB", "400304", 52,
"JJJJ", "404308", 200,
"ARPO", "403898", 2,
"ARPO", "158159", 24,
"BBBBBBB", "700805", 2,
"ARPO S.L.", "900245", 24,
"JJJJ", "501486", 2,
"JJJJ", "400215", 210,
"JJJJ", "504379", 26,
"HARPO", "900222", 400,
"BBBBB", "109700", 46,
"ARPO", "142173", 14,
"BBBBB", "400586", 22,
"ARPO", "401605", 322
)