Away from the mainstream, algorithms fail
Mainstream listeners receive more accurate music recommendations from streaming services than fans of non-mainstream music. This is the result of a study conducted by German, Austrian and Dutch universities.
A team of researchers from Graz University of Technology, Johannes Kepler University Linz, the University of Innsbruck and Utrecht University has investigated the accuracy of algorithm-based music recommendations for listeners of mainstream and non-mainstream music. A data set of the previous listening behavior of 4148 users of the music streaming platform Last.fm was used, half of whom mainly listened to non-mainstream music and the other half mainly listened to mainstream music.
The researchers compared the listening habits of the individual groups and determined which people most frequently listened to music outside their preferred genres and how broadly the music genres listened to were within each group.
Those who mainly listened to music such as ambient were most likely to be willing to listen to music that was actually preferred by hard rock, folk or electronica fans. Those with a preference for energetic music were the least likely to listen to music favored by folk, electronica or ambient followers. Instead, they listened to the widest variety of genres, for example hard rock, punk, singer/songwriter and hip-hop.
Using the computer model, the researchers predicted how likely it was that the different groups of non-mainstream listeners would actually like the recommendations generated by the four common algorithms. The recommendations for lovers of predominantly energetic music seemed to be the least accurate, while they were the most accurate for ambient listeners.
More info:
https://www.know-center.tugraz.at/epj-data-science-algorithmus-basierte-musikempfehlungen/
