CODEWITHSUNDEEP
r
ganong
ganong

Reputation: 311

Using RecordLinkage to add a column with a number for each person

I'd like to do what I think is a very simple operation -- adding a column with a number for each person to a dataset with a list of (potentially) duplicative names. I think that I am close. This code looks at a dataset of names, does pairwise comparisons, and appends a column whether there is a likely match. Now I just want to go one step further -- instead of dropping duplicates, I want to come up with a unique identifier.

Peter

Example:

Peter

Peter

Peter

Connor

Matt

would become

Example:

Peter -- 1

Peter -- 1

Peter -- 1

Connor -- 2

Matt -- 3

library(RecordLinkage)
data(RLdata10000)
rpairs <- compare.dedup(RLdata10000, blockfld = 5)
p=epiWeights(rpairs)
classify <- epiClassify(p,0.7)
summary(classify)
match <- classify$prediction
results <- cbind(classify$pairs,match)

Upvotes: 2

Views: 716

Answers (2)

user5352295
user5352295

Reputation: 51

small rewrite avoiding that the weights and classifier have to be tuned with the IDs,

df_names <- data.frame(Name=c("Peter","Peter","Peter","Connor","Matt"))

df_names %>% compare.dedup() %>%
             epiWeights() %>%
             epiClassify(0.3) %>%
             getPairs(show = "links", single.rows = TRUE) -> matches

left_join(mutate(df_names,ID = 1:nrow(df_names)), 
          select(matches,id1,id2) %>% arrange(id1) %>% filter(!duplicated(id2)), 
          by=c("ID"="id2")) %>%
    mutate(ID = ifelse(is.na(id1), ID, id1) ) %>%
    select(-id1)

Upvotes: 5

ganong
ganong

Reputation: 311

I figured out the answer to my own question.

df_names <- df_names %>% mutate(ID = 1:nrow(df_names))
rpairs <- compare.dedup(df_names)
p=epiWeights(rpairs)
classify <- epiClassify(p,0.83)
summary(classify)
matches <- getPairs(classify, show = "links", single.rows = TRUE)

this code writes an "ID" column that is the same for similar names

matches <- matches %>% arrange(ID.1) %>% filter(!duplicated(ID.2))
df_names$ID_prior <- df_names$ID

merge matching information with the original data

df_names <- left_join(df_names, matches %>% select(ID.1,ID.2), by=c("ID"="ID.2"))

replace matches in ID with the thing they match with from ID.1

df_names$ID <- ifelse(is.na(df_names$ID.1), df_names$ID, df_names$ID.1)

Upvotes: 2

Related Questions