visNetwork

Let’s reload the Game of Thrones network.

library(igraph); library(tidyverse)
edges <- read_csv("../data/GOT-edges.csv")
g <- graph_from_data_frame(d=edges, directed=FALSE)

Next, let’s load in the visNetwork package. Make sure to install if you don’t have it.

Although not required, it’s easiest to create the network using the igraph object (g).

library(visNetwork)
visIgraph(g) %>%
  visNodes(label = V(g)$name) %>% # adds labels of names
  visInteraction(navigationButtons = TRUE) # adds the zoom in and out

Alternatively, we can create the edges into a visNetwork objects: edges and nodes.

One way is to use the as_data_frame() function to pull out each element into its respective object.

# create the links and nodes
edges <- igraph::as_data_frame(g, what="edges")
nodes <- igraph::as_data_frame(g, what="vertices")

# node id must be name as from/to edges 
nodes$id <- nodes$name

Unlike the igraph, when using nodes/edges you can add columns to indicate properties about the nodes or edges.

For example, if you want the labels of the nodes, you can create a new column named label.

nodes$label <- nodes$name

visNetwork(nodes, edges) %>%
  visIgraphLayout() %>% # this forces a visIgraph Layout
  visInteraction(navigationButtons = TRUE)

Let’s instead make the node size relative to betweenness.

First, we need to get betweenness.

b <- betweenness(g) %>%
  sort(decreasing=TRUE)

head(b)
##       Jon    Robert    Tyrion  Daenerys      Robb     Sansa 
## 1279.7534 1165.6025 1101.3850  874.8372  706.5573  705.1986

So Jon, Robert, Tyrion and Daenerys are the “gate-keepers” (highest betweenness).

nodes$size <- ((betweenness(g) / max(betweenness(g)))+.2)*40

visNetwork(nodes, edges) %>%
  visIgraphLayout() %>% # this forces a visIgraph Layout
  visInteraction(navigationButtons = TRUE)

Community Detection

Let’s run community detection using a built-in algorithm in igraph.

clp <- cluster_label_prop(g)

library(RColorBrewer) #install if you don't have it
nodes$color.background <- brewer.pal(12, "Set3")[as.factor(clp$membership)]
nodes$color.border <- "black"
edges$color <- "black"

visNetwork(nodes, edges) %>%
  visIgraphLayout() %>% # this forces a visIgraph Layout
  visInteraction(navigationButtons = TRUE)

Alternatively, we can change the node shape to a circle which will put the labels inside of each node. While this is more aesthetically pleasing, this removes size as proportional to Betweenness.

nodes$shape <- "circle"

visNetwork(nodes, edges) %>%
  visIgraphLayout() %>% # this forces a visIgraph Layout
  visInteraction(navigationButtons = TRUE)

There are a TON of different customizations with visNetwork and igraph.

Also, there are other network packages in R like ggraph, networkD3 and edgebundleR. Explore!

For example…

library(edgebundleR) #install if you don't have it
edgebundle(g, fontsize = 6)