Um modelo para previsão do sucesso no mercado musical

Abstract

The music market is extremely competitive and moves billions of dollars every year. In the United States alone, there are over 1400 music labels currently registered. Standing out in this scenario is an arduous mission. In this dissertation, a model for success prediction in the music market is presented, which artists and record labels may use to direct their efforts into songs with higher potential to return profit. The proposed model makes avail of information about the songs to predict, even before their release, whether they will appear or not in the Top 50 Global ranking from the streaming platform Spotify. To validate this model, we chose as a baseline the most similar model already consolidated in the scientific literature. The baseline employs the kind of information we have employed in our model, but with a distinct approach with respect to the data preparation. Our results are 920% better than those achieved by the baseline.