在R中的定向网络中查找匹配项

Question

I am looking at a directed data set using iGraph based on people who follow one another on Twitter. I have a data table

follower_user_id | followed_user_id | gender_of_follower | gender_of_followed
               1 |                2 |                 F  |                  M 
               2 |                3 |                 M  |                  M
               3 |                2 |                 M  |                  M 

and so on..

I would like to assess the matches, so in which situations do the users mutually follow one another so that I can further look at who was not followed back by anyone and whether men for example are more likely to be followed back than women (etc. ). But I am not sure how to filter down for the matches in the first place.

So far, I think the best way to do this is by using a matrix of all user ids against all user ids and counting the time each pair appears together.

M <- table (df$follower_user_id, df$followed_user_id)
follower.followed.matrix <- M %*% t(M)


XXXXX 1    2   3 
1     0    1   0
2     0    0   1
3     0    1   0 

But I am unsure of how to merge ones where converses are combined (eg where 2-3 pairing = 2) Is it possible to use the 'reshape2' package to melt for a directed network?

I am thinking the best way to do it is in this way and then merging gender data into a new data.table of matches but I am open to suggestions for how to filter for this data in a more efficient way. I am still new to R so any help is appreciated.

Answer 1

Getting the matrix that you need - the adjacency matrix - is built into igraph. I will use a slightly bigger example than yours to make sure that the solution handles all cases.

Data: Enlarged follower Network

FOL = read.table(text="follower_user_id followed_user_id gender_of_follower gender_of_followed
  1  2  F  M 
  2  3  M  M
  3  2  M  M 
  4  2  M  M
  2  4  M  M
  2  5  M  F
  4  5  M  F",
header=TRUE)

## turn it into a graph and compute adjaceny
g = graph_from_edgelist(as.matrix(FOL[,1:2]))
AM = as.matrix(as_adjacency_matrix(g))

Now with the adjacency matrix you can quickly compute the pairs for which A follows B and B follows A.

sapply(1:5, function(x) { AM[x,] * AM[,x] })
     [,1] [,2] [,3] [,4] [,5]
[1,]    0    0    0    0    0
[2,]    0    0    1    1    0
[3,]    0    1    0    0    0
[4,]    0    1    0    0    0
[5,]    0    0    0    0    0

You can see that the needed pairs are (2,3), (2,4), (3,2) and (4,2).