Uses of unsupervised ML on network data

If you take a look at the Karate Club website, you will probably notice that the two approaches to unsupervised ML fall into two categories: identifying communities or creating embeddings. Unsupervised ML can be useful for creating embeddings not just for nodes, but also for edges or for whole graphs.

Community detection

Community detection is the easiest to understand. The goal of using a community detection algorithm is to identify the communities of nodes that exist in a network. You can think of communities as clusters or clumps of nodes that interact with each other in some way. In social network analysis, this is called community detection, because it is literally about identifying communities in a social network. However, community detection can be useful outside of social network analysis involving people. Maybe it helps to think of a graph as just a social network of things that somehow interact. Websites interact. Countries and...