Catching the Earworm: Understanding Streaming Music Popularity Using Machine Learning Models

Shanghai High School International Division Shanghai, China

Abstract

The digitization of music has fundamentally changed the consumption patterns of music, such that the music popularity has been redefined in the streaming era. Still, production of hit music that capture the lion’s share of music consumption remains the central focus of business operations in music industry. This paper investigates the underlying mechanism that drives music popularity on a popular streaming platform. This research uses machine learning models to examine the predictability of music popularity in terms of its embedded information: audio features and artists. This paper further complements the predictive model by introducing interpretable model to identify what features significantly affect popularity. Results find that machine learning models are capable of making highly accurate predictions, suggesting opportunities in music production that could largely increase possibility of success within streaming era. The findings have important economic implications for music industry to produce and promote the music using controllable and predictable features tailored to the consumer’s preferences.