Music intelligence: Granular data and prediction of top ten hit songs

摘要

In the music market, superstars significantly dominate the market share, while predicting the top hit songs is notoriously difficult. The music intelligence technology, retrieving and utilizing granular acoustic features of songs, provides opportunities to improve the prediction of top hit songs. Using data on 6209 unique songs that appeared in the weekly Billboard Hot 100 charts from 1998 to 2016, especially acoustic features provided by Spotify, we investigate empirically how the top-10-hit-songs likelihood prediction is improved by acoustic features. We find that some acoustic features (e.g., danceability, happiness, and some metrics of timbre and pitch) significantly improve the model's ability to predict the top-10-hit-songs probability. These results suggest that the granular data, provided by the music intelligence technology, carries a substantial predictive value in the era of online music streaming.