Online Music Preferences of Different Generations: an Acoustic Perspective

The relationship between music preferences with age is an interesting but rarely
studied topic. The limited relevant research mostly concentrated on the musical
genre level or a certain age group. In this thesis, we study the music preferences
of different generations from an acoustic perspective, intending to understand the
differences and trends in the various music tastes. We jointly took advantage of the
LFM-1b dataset, MusicBrainz, and AcousticBrainz to collect the user’s listening
events and acoustic information. F-test feature selection and support vector machines
were then used to analyze the correlation between generations’ preferences
and acoustic features. We finally found the most relevant acoustic features and revealed
their trends along with the generation changing. For example, the younger
listeners tend to listen to songs that are spectrally more complex, while the older
listeners show more interest in “sad” music. We further interpreted these features,
attempting to understand the reasons for these trends. Our findings suggested that
some previously observed genre preferences of relevant work were associated with
these acoustic features. Therefore, this thesis provides a new perspective for the
study on music preferences and age.