We love rock 'n' roll: analyzing and predicting friendship links in Last.fm

  • Authors:
  • Kerstin Bischoff

  • Affiliations:
  • L3S Research Center/Leibniz Universität Hannover, Appelstrasse, Hannover, Germany

  • Venue:
  • Proceedings of the 3rd Annual ACM Web Science Conference
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

Music plays an important role in our everyday lives. Not surprisingly, shared musical taste is said to lead to social attraction. In this paper, we study in detail friendship links on the social music platform Last.fm, asking for similarities in taste as well as on demographic attributes and local network structure. On Last.fm, users connect to 'online' friends as usual, but also indicate strong 'real-life' friends by co-attending the same events. Thus, we can contrast these online ties with offline links of different strength. Complementing the analysis, we learn to predict both kinds of ties automatically, including public interaction data as additional relevant features. Our results emphasize the predictive power of the simple measure of mutual friends, while the indicative value of similarity on taste (though increasing with tie strength) is negligible.