Uncovering Friend Groups in Social Networks

|Jan 11th, 2022|

Social network analysis is a multi-billion dollar business. By clustering online connections into groups of friends, tech organizations can better target their tools towards shared patterns of social behavior. In this talk, we’ll discuss a graph clustering technique called Markov Clustering. The technique is closely related to the PageRank centrality algorithm, which was invented by the Google founders to provide superior website rankings. We will derive both PageRank and the Markov Clustering algorithm based on a series of experiments. These experiments will be carried out using Python’s NetworkX graph analysis library. What to Expect: We will leverage Python to explore a series of real and simulated social networks. During our explorations, we’ll show how friend groups can be easily identified by simulating random traffic flow across the networks. We’ll also demonstrate how traffic flow probabilities can be accurately computed using simple matrix multiplications. Eventually, we’ll utilize these multiplications to code-up our simple yet elegant friend-clustering algorithm.