We Know What You Did Last Sonar: Inferring Preference in Music from Mobility Data

Digital music platforms allow us to collect feedback from users in the form of ratings. This type of information is explicit about the users musical preferences. In contrast, intrinsic feedback provides contextual information from which preferences can be inferred, some quite obvious like amount of playing activity, playlists, and others less direct like activity in or components of social networks. Here we focus on physical intrinsic feedback in the form of mobility traces on a music festival with multiple stages, and analyze it to infer music preferences. To the best of our knowledge, this is the first research work that exploits physical contextual clues and human mobility behavior from WiFi traces to approximate ratings that are later used to: (1) measure musical similarity among musical bands, and (2) estimate the effect of loyalty of a physical audience (i.e. going further than measuring sheer number of attendees). As part of this work, we developed a novel metric to measure weighted user rating from the mobility patterns of individuals during a music festival, incorporating physical contextual data in addition to factors commonly used in ranking systems. The experiments reveal groups of people with similar musical preference, and adjusted rankings that identify acts that, even if small, were highly successful in terms of their effect on the audience.