Predicting Song Popularity: Universal Models From International and Inter-Genre Chart Sources
Every year, a large number of new songs is produced and released. These songs compete against each other for disc sales, online stream counts, etc. In order to quantify which songs perform best, various sources regularly publish music charts. Such charts are then often used to construct models for the popularity of songs since the most popular songs naturally rank highest in the charts.
However, most approaches to this are limited in that they only use a single chart source and thus do not capture a universal (i.e., international, inter-genre) picture of popularity. The goal of this thesis is to analyze the data of various international, inter-genre chart sources and to test whether and how such sources can be combined to produce popularity models that perform well universally.