Visulization and clustering Facebook Ego Network in R (part 1)

By giving people the power to share, we're making the world more transparent
Mark Zuckerberg

Mining social network is an interesting task in the world of facebook, twitter, google plus and other online social networking services. 

I this post I will use NetworkD3 to visualize friendship in facebook ego network and use Markov clustering algorithm to cluster these ego-networks. The data set is taken from snap project

Noteego network (aka. personal network) is a network of connections between someone's friends. You can find more information about ego network here


So let's take a look at data set. Each ego-network we have circlesedgesegofeatfeatfeatnames file. The detail of these file is described here. In this post, I'm going to cluster people only by their friendship, so we just focus in the edges file

  • edges : The edges in the ego network for the node 'nodeId'. Each line represent a friendship between two users.

For example, the third line of 0.edges file

 24 346

means user 24 and user 346 are friends.

Our very first ego-network visualization

To visualiza ego network, I use networkD3, it's very fast, beautiful and importantly, it can be interactable.

We load data from 0.edges file

 graph.edges <- read.csv(
  file = 'data/facebook/0.edges',
  sep=" ", header=F,
  col.names=c("source", "target"))
graph.edges <- graph.edges - min(graph.edges) + 1
graph.nodes <- data.frame(
  id=seq(max(graph.edges) - min(graph.edges) + 1))
graph.nodes$group <- 1

And then we create forceNetwork. Notice that index of Links in forceNetwork is start from 0, while index of edges in 0.edges file is start from 1, so we must subtract 1 from edges

 network <- forceNetwork(
  Links = graph.edges - 1, Nodes = graph.nodes,
  Source = "source", Target = "target",
  NodeID = "id",
  Group = "group", opacity = 0.8)

And this is the result

Looking at this graph, we see some cluster here, and I find an algorithm to do it for me.

Clustering with MCL

We build a function graph.cluster to cluster a graph with input are nodes and edges. In this function, we first create an adjacency matrix adjacency and run mcl algorithm to this matrix to find our cluster

Next we use this function to reassign group to graph.node data frame


graph.cluster <- function(nodes, edges){
  n <- nrow(graph.nodes)
  adjacency <- matrix(0, n, n)
  index <- cbind(
    graph.edges$source, graph.edges$target)
  adjacency[index] <- 1
  cluster <- mcl(
    x = adjacency, addLoops=TRUE, ESM = TRUE)

graph.nodes$group <- graph.cluster(
  graph.nodes, graph.edges)

Finally, we use networkD3 to create graph

 network.cluster <- forceNetwork(
  Links = graph.edges - 1, Nodes = graph.nodes,
  Source = "source", Target = "target",
  NodeID = "id",
  Group = "group", opacity = 0.8)

This code generate a graph like this below image

Now we see clusters of this ego network more clearly, there are a big orange cluster in the bottom of graph which contains many nodes, some smaller clusters with only 7-10 node like a light blue cluster in top of graph, some pink cluster, and there are many cluster which contains only one node, which is cluster for people has no friends in this ego network.


It's my very simple approach to cluster ego-network. This method clusters ego-network with only frienship information. MCL works very fast with ego-network with about 1000 nodes. You can try by yourself with larger network (simply change input file to another edges file)

With real data, here is my clustered go-network

The result is very interesting, the purple cluster are my friends from my university, the green one are my friends from my secondary and high schools. the blue one are my lover's friends and so on.

So we have a tool to discover an ego-network quickly in this post. I hope you find fun with this. In the next posts, I will try some more complicated method which cluster ego network with more information of people. See you there.

You can find full code and data of this project in


2 thoughts on “Visulization and clustering Facebook Ego Network in R (part 1)

  1. amar says:

    Thanks for the post. can you please provide me the link for the next post you talked about.
    (more complicated method which cluster ego network with more information of people)?


    Liked by 1 person

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s