Top-N music recommendation framework for precision and novelty under diversity group size and similarity

The growth of music streaming market is expected to be boosted by the rising use of smart devices and the streaming platforms in the forecast period. Music recommendation systems for groups have been intensively studied as the application scenario becomes more dynamic, where different kinds of group formulations lead to different levels of similarity in music tastes. This paper proposes a music recommendation framework to generate a Top-N list in high-similar, high-dissimilar, and regular groups. Additionally, we investigate the performance of the recommendation system in a wide range of aspects, including precision, ranking quality, and novelty. Based on the listening frequency of each user and for each track, implicit feedback is also derived from the track popularity. To prevent popularity bias, we combine track popularity and group popularity into latent factor models to improve the performance of the recommendation algorithm. As a result of our evaluation, we found that the proposed methods provided excellent recommendations regarding accuracy, diversity, and novelty.