Comfortably Numbers

Philippe Haldermans, Jonas Kiesekoms and Maarten Elen of PXL-Music Research investigate how data science influences the music industry.



Together with Ancienne Belgique, Clubcircuit and Botanique the team conducts research on data science in the music industry. They collected data of more than half a million songs and more than 7000 concerts in Belgium. With these figures the researches now investigate if the success of future shows can be predicted. Other research questions are: which genres are popular in Belgium? Does the audience consume more beer at metal shows or at jazz concerts? Do frontwomen attract a greater audience than frontmen? These and many other questions form the basis for a predictive model. The ultimate goal: to give young experimental bands more opportunities to perform. If organisations can make a better risk analysis they can take the freedom to experiment artistically.

Read more about the research here.


All work and no play makes Jack a dull boy


To tackle the endless discussion on which record ought to be played in their lab, these guys developed a nice visualisation tool. So they could find the music that matched the emotions and energy involved in their research. They connected to the Spotify API and analysed music according to its mood and energy.

To improve their playlists streaming service Spotify collects a great amount of data from each song. PXL-Music Research dived into these data. Two parameters jumped out: Mood (how positive or negative is the song?) and Energy (is there a lot of energy in the song?). These two parameters can form two axes and a field with four zones (or quadrants) in which each song can be placed: tender (positive mood, low energy), uplifting (positive mood and high energy), sad (negative mood, low energy) and angry (negative mood, high energy):

Let us illustrate this with a few examples. Slayer clearly makes angry music:



James Blake mostly wanders in the sad segment:



Beck is a chameleon. His songs are all over the place: