Predicting Music Popularity Patterns based on Musical Complexity and Early Stage Popularity

Abstract

This paper investigates the problem of predicting popularity of music. In particular, we consider musical complexity as a cue that can be extracted from the audio signal and used for popularity prediction. In addition, we examine the effectiveness of the early stage popularity for long-term popularity prediction. We formulate the popularity prediction problem as a classification problem predicting popularity evolution patterns in a music ranking chart, such as the highest rank of a song over the whole time period, the growth/declination rate in the chart, the duration for which the song appears in the chart, etc. We conduct an experiment with the data collected from the Billboard Rock Songs Chart for about five years. It is found that the two types of features are effective for predicting popularity patterns when used together.