Identifying time zones in a large dataset of music listening logs

Gabriel Vigliensoni, Ichiro Fujinaga

Abstract: Knowing where listeners are is an important contextual dimension that can be used in context-aware music recommendation systems to improve their performance. This paper presents our research on identifying the time zone where listeners are by analysing their weekly aggregated music listening profiles. We collected a large dataset of full music listening histories (N=594K) of users of the Last.fm's scrobbler service from all around the globe, and formulated six approaches for identifying the time zone where these listening profiles have been generated based on their listeners' behaviour. The performance of these approaches was compared with a manually labelled dataset of listening profiles' time zones. We found that the best method was based on the assumption that people, in general, sleep during night time and submit fewer music logs. This approach, implemented by estimating the local minima of people's weekly aggregated listening profile, resulted in a 75 percent correctly identified time zones with a tolerance of +/- 1 hour.