Detecting the next pop star: Artist breakthrough predictions based on listener characteristics

Nowadays, large-scale music streaming provides rich insights into listening activities, listener profiles, preferred genres, similar listeners and social networks. This information opens the door to a new approach towards future star detection. This paper proposes a model which detects musical trendsetters, based on listener data from the music database ‘Last.fm’, over a ten-year period. Each user is rated in terms of how often he or she listened to an artist before that artist broke through: The user’s trendsetting score. It is studied what characterizes the most influential trendsetters: Their age, Last.fm membership, openness to novelty, music originality and/or network strength? Based on the strongest indicators of being a trendsetter, a ‘trendsetter detection model’ and a ‘trendsetter profile’ are built. These models classify users into ‘trendsetters’ and ‘non-trendsetters’. Based on the variables included in the trendsetter detection model, the ‘star prediction model’ is proposed. This model analyses an artist’s listener base characteristics to determine whether that artist’s listeners fit into the trendsetter profile. Based on this information, the model predicts which musical talents will break through. Various stakeholders in the music industry can use this model to target those artists with the most promising career perspective.