Inferring personal traits from music listening history

Nowadays, we often leave our personal information on the Internet without noticing it. People could learn things about you from these information. It has been reported that it is possible to infer some personal information from the web browsing records or from blog articles. As the music streaming services become increasingly popular, the music listening history of one person could be acquired easily. This paper investigates the possibility for a computer to automatically infer personal traits such as gender and age from the music listening history. Specifically, we consider three types of features for building the machine learning models, including 1) statistics of the listening timestamps, 2) song/artist metadata, and 3) song signal features, and evaluate the accuracy of binary age classification and gender classification utilizing a 1K-user dataset obtained from the online music service Last.fm. Our study brings about new insights into the human behavior of music listening, but also raises concern over the privacy issues involved in music streaming services.