Predicting Hit Songs With 97% Accuracy

This article examines the difficulty of predicting hit songs and how machine learning can be used to increase accuracy. It discusses the traditional approach of measuring song elements from large databases to identify lyrical aspects of hits, and how a different methodological approach of measuring neurophysiologic responses to a set of songs provided by a streaming music service can be used to identify hits with greater accuracy. The article also discusses how applying machine learning to the neural response to the first minute of songs can accurately classify hits 82% of the time.