Empirical Analysis of Track Selection and Ordering in Electronic Dance Music using Audio Feature Extraction

Disc jockeys are in some ways the ultimate experts at selecting and playing recorded music for an audience, especially in the context of dance music. In this work, we empirically investigate factors affecting track selection and ordering using mixes created for the Essential Mix. The Essential Mix is a well known weekly radio show on BBC Radio 1 that showcases various styles of electronic dance music. We use automatic content-based analysis and discuss the implications of our findings to playlist generation and ordering. Timbre appears to be an important factor when selecting tracks and ordering tracks, and track order itself matters, as shown by statistically significant differences in the transitions between the original order and a shuffled version. We also apply this analysis to ordering heuristics and suggest that the standard playlist generation model of returning tracks in order of decreasing similarity to the initial track may not be optimal, at least in the context of track ordering for electronic dance music.