Group by two columns and union of levels in R

问题

I am stuck on a problem that seem trivial but I am unable to figure it out right now. I don't even know how to formulate it properly, if you have any suggestions, you are welcome. I have a data.frame which I want to group/index depending on two columns. The thing is, the rows I want to group do not share the same values in those columns. Rather, some rows have the same value in one column, and then some of those rows have a common value with different rows in the second column (which I also want to include in the grouping). Here is a minimal example, I hope this makes it a bit clearer:

id V1 V2 group_id
1   a  c        1
2   a  d        1
3   b  d        1
4   w  y        2
5   w  z        2
6   x  z        2

Rows 1 and 2 have the value a of column V1 in common. But I not only want to group them, but also row 3, which is "connected" via the value d of column V2. Right now, I am only able to group rows 1,2 and 2,3 separately.

The same is true for the 2nd group, here I want to group values with either w in V1 or z in V2. x and y are irrelevant.

Any help is highly appreciated.

回答1:

Here's how you could do that with the cluster function from the igraph package:

library(igraph)
relations <- data.frame(from=df$V1,to=df$V2)
g <- graph_from_data_frame(relations)
group_id <- data.frame(V=names(clusters(g)$membership),
                       cluster=clusters(g)$membership,stringsAsFactors=FALSE)
left_join(df,group_id,by=c("V1"="V"))

  id V1 V2 group_id cluster
1  1  a  c        1       1
2  2  a  d        1       1
3  3  b  d        1       1
4  4  w  y        2       2
5  5  w  z        2       2
6  6  x  z        2       2

来源：https://stackoverflow.com/questions/43482086/group-by-two-columns-and-union-of-levels-in-r