IJMSTA - Vol. 1 - Issue 2 - September 2019
A Smart Shuffle approach to Playlist Shuffling with User Defined Constraints
Authors: S. Ramroach, P. Hosein
Abstract - We consider the problem of track sequencing for a given music playlist. We assume that a user chooses a set of desirable songs to form a playlist as would be done in applications such as iTunes or Google Play Music. However, instead of using the typical random shuffle feature, we introduce what we call a smart shuffle option in which the user specifies various constraints that must be satisfied when determining the playback sequence. These constraints are based on several attributes of the songs. If the user does not provide any constraints, all attributes are considered equal. The general computational problem is the Travelling Salesman Problem in Euclidean space. We consider the following approaches: three hierarchical clustering techniques, K-means clustering, a greedy swapping approach, and an approximation approach (Christofides' 3/2 -approximation). We then compare performances based on a defined performance metric. We also perform a subjective evaluation to ensure that the proposed model enhances the listening experience of a user.
Keywords: Playlist shuffling, Music sequencing, Optimization, Data Analytics