Classification of Following Intentions Using Multi-layer Motif Analysis of Communication Density and Symmetry Among Users

The information diffusion on social media shows no signs of stopping, and for many marketers, having influencers spread information has become a common advertising method. However, the use of social media has diversified, and the intentions behind following other users can vary greatly. Correspondingly, there are several patterns of communication on social media, and based on the density and symmetry of these communications, it is believed that one can infer the intentions of the users who follow others. In this study, we consider retweets, replies, and mentions, three types of communication in the context of follower relationships on Twitter, as a multi-layer graph. We propose multi-layer motifs by categorizing the edges, and we associate motif patterns with follower intentions to infer users’ follow intents. Through experiments using real data, we confirm that our proposed multi-layer motifs can extract link patterns leading to follow intentions that would not be detectable using traditional single-layer motifs.