Clustering of Twitter Networks Based on Users’ Structural Profile

Twitter’s ability to connect users around a given topic provides an insight into the complex mechanisms that grant positions of influence to a subset of users. This work focuses on clustering a collection of Twitter topic networks through an interpretable approach centered on the asymmetric relations on the platform. We create a network representation based on directed graphlet-orbits, using graphlets with 2–4 nodes. Our method has two main steps; first, we identify structural profiles for the network users. Then, we create network embeddings using the previous profiles and establish groups within the collection. We show the applicability of the proposed method by analyzing 50 real networks generated around trending topics in Mexico and discussing the identified user profiles from the viewpoint of the social power dynamics they reflect.