Last year, Coursera offered a Social Network Analysis course where the instructor is one of the Computational Social Scientist of Facebook, Lada Adamic. Too bad I did not finish this course because it overlapped with another course I was finishing. When this course is offered again, I will definitely complete this since I have to visualize more graphs that deals with the network of textual data using association rule learning.
So here’s my Facebook social network. I wanted to mine more information about my friends but after Facebook retired Oauth 1.0, it really restricted a lot of information that you can mine from your friends. I guess that’s a good thing because of privacy information concerns. So I can only mine the network map though I found a way to mine my textual posts through the archive feature but I have to build some program, maybe in C++ or Python, to extract the data properly.
The size of each label is based on the mixed degree, meaning our mutual friends. The colors of each line the intensity, the purple being the highest, green is middle and blue is the least.
|Facebook Social Network|
It is fun to see how my friends (nodes) are clustered together based on their distance of connectivity and how each single friend can be interconnected with two clusters, putting their spot in between as it is the average distance between two clusters.
This is not really a particular big network, consisting of 2,500+ nodes (friends) and 55,000+ edges (how each node is connected). Although with a pretty decent PC running at 8GB RAM, Gephi was already performing slow in running ForceAtlas 2.
See the upper left sub giant network? I probably need to clean up my friends list because they do not make sense. These are my Facebook friends who are closely related to each other by using the recommended friend list of Facebook – a behavioral pattern where a group of men who looks at the stance of your half-naked pictures.
I wonder if there’s any platform where I can upload my big network so you can interactively see the nodes and labels.