在加入 Dplyr 之前，使用 RecordLinkage 匹配两个数据框中的不等名称列

Question

以下是我想使用 Dplyr 按名字和姓氏加入的两个数据框。

full_join(Names1,Names2,by=c("FirstNames",LastNames")

但是，某些名称的拼写和格式存在差异。我想使用 RecordLinkage （或 adist）来匹配相似的名称，然后覆盖名称，以便它们在加入之前在两个数据帧中匹配。 （我也意识到“FirstNames”和“LastNames”列在两个数据框中都有不同的标题......所以必须更改。）

FirstNames<-c("Chris","Shintaro","Doug","Elsa","Bubbles","Kelly","Christine")
LastNames<-c("MacDougall","Yamazaki","Shapiro","Elizabeth Ray","Murphy","Anderson","Yamaguchi")
Pets<-c("Cat","Dog","Cat","Dog","Cat","Snake","Eagle")
Names1<-data.frame(FirstNames,LastNames,Pets)

FirstNames2<-c("Chris","Doug","Shintaro","Bubbles","Elsa")
LastNames2<-c("MacDougal","Shapiro","Yamazaku","Murphy","Elizabeth")
Dwelling<-c("House","House","Apartment","Condo","House")
Names2<-data.frame(FirstNames2,LastNames2,Dwelling)

以下是我在 RecordLinkage 包中使用的一些步骤，但我遇到了“错误：所有 select() 输入必须解析为整数列位置。”这是因为数据框的行数不同吗？

Results <- compare.linkage(Names1, Names2, blockfld = 1, strcmp = T, exclude = 3)

PairsSelect <- 
Results$pairs %>% 
select(firstNameSim = FirstNames, lastNameSim = LastNames)

我想继续使用代码将名称列与上面的 PairsSelect 数据连接起来，以便我可以在名字和姓氏列中找到匹配项，然后覆盖两个数据框，使每个数据框具有相同的名称拼写和格式在加入之前。我只是不确定如何最好地处理一个数据帧（Names1）比另一个数据帧（Names2）包含更多行的事实。

任何关于如何实现这一目标的指导将不胜感激！

Answer 1

我已将键与进入“compare.linkage”函数的数据分开，这与数据库世界观相比有点狡猾，但只要你在加入之前不重新排列相对于 df 的名称compare.linkage() 流回来的一切都将保持正确。

library(dplyr)
library(RecordLinkage)

FirstNames<-c("Chris","Shintaro","Doug","Elsa","Bubbles","Kelly","Christine","Thomas","George")
LastNames<-c("MacDougall","Yamazaki","Shapiro","Elizabeth Ray","Murphy","Anderson","Yamaguchi","DelTorre","Fibonacci")
Pets<-c("Cat","Dog","Cat","Dog","Cat","Snake","Eagle","Shark","Parrot")
Names1<-data.frame(FirstNames,LastNames,Pets,FullNames = paste0(FirstNames, " ", LastNames))

FirstNames2<-c("Chris","Doug","Shintaro","Bubbles","Elsa","George")
LastNames2<-c("MacDougal","Shapiro","Yamazaku","Murphy","Elizabeth","Fibos")
Dwelling<-c("House","House","Apartment","Condo","House","Camper")
Names2<-data.frame(FirstNames2,LastNames2,Dwelling,FullNames2 = paste0(FirstNames2, " ", LastNames2))

df1 <- mutate(id1 = row_number(), Names1)
df2 <- mutate(id2 = row_number(), Names2)

Results <- compare.linkage(Names1, Names2, strcmp = T, exclude = c(1:3))

BestResults <- 
    Results$pairs %>% 
    arrange(desc(FullNames))

BestResultsJoined<- 
    full_join(BestResults,df1,by="id1")%>% 
    full_join(df2,by="id2")%>% 
    select(FullNamesSim=FullNames.x,FullNames1=FullNames.y,FullNames2,id1,id2)
BestResultsJoined

这将为您提供最接近匹配到最不接近匹配的排序列表。看起来您不想考虑任何小于 0.95 的相似度（按 Jaro-Winkler 距离）。