Chapter 14 Chapter 14: Network visualization in igraph

This chapter aims to introduce you the basics of network visualization. Note that for a very large network (> 100 nodes) it becomes very difficult to create good visualizations using basic igraph plotting functions. You may need to venture to other R packages (see Chapter 15 - under development). A nice open-source software for creating great network visualizations is Gephi.

14.1 Set up

We will use the karate network from Chapter 5, which contains various node and edge attributes.

library(igraph)
library(igraphdata)
library(tidyverse)

data('karate')

# import your node attributes 
node_info <- read.csv('data/karate_nodes_added.csv', header = T)

# very important to ensure that node order is identical 
node_info <- node_info %>% arrange(factor(node, levels = V(karate)$name))
## This graph was created by an old(er) igraph version.
## ℹ Call `igraph::upgrade_graph()` on it to use with the current igraph version.
## For now we convert it on the fly...
identical(V(karate)$name, node_info$node) # sanity check that the node name order is identical
## [1] TRUE
# add the 'gender' attribute
karate <- set_vertex_attr(karate, 
                          name = 'gender',
                          value = node_info$gender)

# add the 'belt' attribute
karate <- set_vertex_attr(karate, 
                          name = 'belt',
                          value = node_info$belt)

# add the 'age' attribute 
karate <- set_vertex_attr(karate, 
                          name = 'age',
                          value = node_info$age)

# initialize all edges with the same label 
E(karate)$edge_type <- 'same' 

# re-assign those with mixed edges to a new label 
E(karate)$edge_type[E(karate)[V(karate)[V(karate)$gender == 'male'] %--% V(karate)[V(karate)$gender == 'female']]] <- 'different'

summary(karate) # there should be a new edge type 
## IGRAPH 4b458a1 UNW- 34 78 -- Zachary's karate club network
## + attr: name (g/c), Citation (g/c), Author (g/c), Faction (v/n), name (v/c), label (v/c), color (v/n), gender (v/c), belt (v/c), age (v/n), weight (e/n), edge_type (e/c)

14.2 Visualizing node attributes

14.2.1 Color nodes by their categorical attributes

# color nodes by their attributes (categorical)
plot(karate, vertex.color = factor( V(karate)$gender )) # notice the use of factor() 

# plot(karate, vertex.color = V(karate)$gender) # this throws an error

If you would like to specify your own colors, you can refer to this website for the names of colors in R and assign them to each category as shown in the following code:

factor( V(karate)$gender ) |> levels() # order of levels is female male
## [1] "female" "male"
my_colors <- c('firebrick', 'slateblue') # firebrick = female, slateblue = male

# use of square brackets to map the colors onto the labels 
plot(karate, vertex.color = my_colors[ factor( V(karate)$gender ) ] ,
     vertex.label.color = 'white' # to make the labels visible
     )

Can you try replacing the gender attribute with the belt attribute?

14.2.2 Color nodes by their continuous attributes

# color nodes by their attributes (continuous) 

# a function to generate color gradients from a pre-specified palette 
palf <- colorRampPalette(c("red", "green")) # smaller values = white 

plot(x=10:1, y=1:10, pch=19, cex=5, col=palf(10))

plot(karate, vertex.color = palf(10)[cut(V(karate)$age, 10)]) # notice the use of cut() 

# cut() chops up the age distribution into 10 bins, each bin is assigned to the color gradient based on the order 
# the number of bins and color bins must be the same

14.2.3 Node shapes by their categorical attributes

# give categorical attributes different shapes 
plot(karate, 
     vertex.shape = c('circle', 'square')[factor(V(karate)$gender)], # notice the use of factor() with gender to subset a vector of pre-defined node shapes
     vertex.color = 'white') 

factor(V(karate)$gender) |> levels() # female/circle, male/square
## [1] "female" "male"

14.2.4 Node size by their continuous attributes

# give continuous attributes different sizes 

plot(karate, vertex.size = V(karate)$age) # notice that the sizes are not very well scaled 

plot(karate, vertex.size = 0.5*V(karate)$age) # you will have to play around to find a suitable scale

14.2.5 Example 1: Color nodes by their community membership

set.seed(1)
community_result <- cluster_louvain(karate)

V(karate)$community <- community_result$membership

plot(karate, vertex.color = V(karate)$community)

14.2.6 Example 2: Size nodes by their degree

plot(karate, vertex.size = degree(karate), vertex.color = 'black') # the scaling issue may apply here

14.3 Visualizing edge attributes

14.3.1 Color edges by their categorical attributes

plot(karate, vertex.color = 'white', edge.color = factor(E(karate)$edge_type)) # note the use of factor()

plot(karate, vertex.color = 'white', edge.color = c('red', 'blue')[factor(E(karate)$edge_type)]) # specifying your own colors 

factor(E(karate)$edge_type) |> levels() # this tells you the ordering of the levels so you can tell that red = different and blue = same 
## [1] "different" "same"

14.3.2 Edge thickness based on their continuous attributes

plot(karate, vertex.color = 'white', edge.width = E(karate)$weight)

14.3.3 Arrows for directed edges

plot(as.directed(karate, mode = 'random'), vertex.color = 'white', edge.arrow.width = 0.5, edge.arrow.size = 0.5)

14.4 Other

14.4.1 Graph layouts

It is good practice to set a seed before executing the graph layout functions, so that you are able to “save” the output and replicate it in the future. Notice that the layout function takes the network as its argument and outputs a 2 column matrix which is the x- and y-coordinates for each of the nodes in the network (explore the my_layout1 R object). Then you can insert that as the input to the layout parameter when you want to plot the network.

## saving a fixed layout for future use 
set.seed(1)
my_layout1 <- layout_in_circle(karate)

set.seed(2)
my_layout2 <- layout_with_dh(karate)

plot(karate, layout = my_layout1)

plot(karate, layout = my_layout2)

See https://kateto.net/network-visualization (Section 4.2 Network layouts) to learn more about the different plotting formats available in igraph.

14.4.2 Plot title

## add a plot title ----
plot(karate, vertex.color = V(karate)$community, main = 'Communities in the karate network')

14.4.3 Add a subtitle

## add a subtitle ----
plot(karate, vertex.color = V(karate)$community, main = 'Communities in the karate network',
     sub = 'The Louvain community detection method was used.')

14.4.4 Multiple plots

## multiple plots ----

par(mar=c(0,0,0,0)+1, mfrow = c(1,2)) # adjust margins, 1 row, 2 columns  

plot(karate, layout = my_layout1, main = 'layout 1')
plot(karate, layout = my_layout2, main = 'layout 2')