17

My general problem is that I loose the vertex names / labels (not sure about the right word here) when generating a graph using iGraph.

I have an edge list IC_edge_sub of a bipartite network, that looks like the following:

  new_individualID new_companyID
1             <NA>     10024354c
3        10069415i      2020225c
4        10069415i     16020347c
5        10069272i      2020225c
6        10069272i     16020347c
7        10069274i      2020225c

I then create a graph element:

IC_projected_graphs <- bipartite.projection(IC_twomode, types = 
                         is.bipartite(IC_twomode)$type)

Then collapse it to identify only connections between companyIDs

IC_projected_graphs <- bipartite.projection(IC_twomode, types =   
                         is.bipartite(IC_twomode)$type)

And then get the adjacency matrix:

CC_matrix_IC_based <- get.adjacency(CC_graph_IC_based); CC_matrix_IC_based

In iGraph node numbering starts at zero and thus also the matrix naming starts at zero. However, I would instead now need the "new_companyID" as specified in the 2nd column of the edgelist in the eventual CC_matrix_IC_based matrix.

Can you help me how to use the information form the original edgelist to put in rownames and colnames in the eventual adjacency matrix?

I googled it and searched stack flow, but could not really find a working answer. Thanks a lot for your help

Andrie
  • 163,419
  • 39
  • 422
  • 472
Henning Piezunka
  • 235
  • 1
  • 3
  • 9

2 Answers2

24

Vertex names are usually stored in a vertex attribute named name in igraph. So, if your graph is stored in the variable g, then you can use V(g)$name to retrieve the names of all the vertices.

Tamás
  • 44,085
  • 11
  • 94
  • 119
2

I know, fairly presumptuous to answer one's own questions.

I think I have resolved it. The key issue was that I had not saved the names when generating the graph. Thanks to Tamas. Without her answer I would not have realized that. Afterwards I needed to ensure not to loose the data. In the following the overall solution:

  # Subsetting / triangulating data for selected games
      GC_edge_sub <- subset (GC_edge, mb_titleID %in% loggames_yearly_sample$mb_titleID)
      GC_edge_sub <- subset(GC_edge_sub, select=c("new_titleID", "new_companyID"))
      head(GC_edge_sub)

  # Generating the vertex names
      vertex_new_companyID <- data.frame(names = unique(GC_edge_sub$new_companyID))
      vertex_new_titleID <- data.frame(names = unique(GC_edge_sub$new_titleID))
      vertex <- rbind(vertex_new_companyID,vertex_new_titleID)

  # Creation of GC_twomode
    GC_twomode <- graph.data.frame(GC_edge_sub, vertices = vertex)
    GC_projected_graphs <- bipartite.projection(GC_twomode, types = is.bipartite(GC_twomode)$type)
    GC_matrix_GC_based <- get.adjacency(GC_twomode)
    dim(GC_matrix_GC_based)

  # Collapsing the matrix
      # Be aware that if you use the classical command # CC_graph_GC_based <- GC_projected_graphs$proj2 it collapses, but looses the colnames and rownames
      # I thus a) create a subset of the adjacency matrix and b) create the lookef for matrix by multiplication    
        rowtokeep <- match(vertex_new_companyID$names,colnames(GC_matrix_GC_based))
        coltokeep <- match(vertex_new_titleID$names,rownames(GC_matrix_GC_based))
        GC_matrix_GC_based_redux <- GC_matrix_GC_based[rowtokeep,coltokeep]
    # We now have a CG matrix.Let's build from this a GG matrix.
        CC <- GC_matrix_GC_based_redux %*% t(GC_matrix_GC_based_redux)
talonmies
  • 67,081
  • 33
  • 170
  • 244
Henning Piezunka
  • 235
  • 1
  • 3
  • 9