A comparison study on familiarity-based and similarity-based social networks

User-based collaborative filtering recommends items to a user based on those items similar users have purchased. User similarity can be defined by their taste similarity or by their friendship. This study compares two social networks, one constructed from users with similar tastes, the other constructed from users' contact lists. Results show that, from the macroscopic perspective, all three networks have small world property. The correlations between different centralization measures of the co-reading networks are stronger than those of the friendship network. From the microscopic perspective, the top users in different centrality measures are consistent for the co-reading networks, but not for the contactship network. From the local perspective, the 3-block model is the best for the contactship network. No correlation exists between users' location and their network positions. Finally, users' contacts and their interest sharers do not overlap.