Meta-path based social event recommendation

Event based social networks (EBSNs) provides users with online to offline social services. Users can organize events online and participate in them offline. Particularly, social event recommendation is a hot issue in this field. The existing studies have mainly focused on factors-based or graph-based social event recommendation, which can hardly exploit the semantic information in the social networks thoroughly. To address this problem, a meta-path based social event recommendation method is proposed. The meta-path contains rich semantic information, being an important feature of the heterogeneous information networks. Firstly, a heterogeneous network is constructed, and then various candidate meta-paths are utilized to calculate the correlation matrix between the users and events. On this basis, the weight of each meta-path is solved by optimization method. Accordingly, the social event recommendation is realized in terms of the linear combination of correlation matrix and meta-path weights. The algorithm is validated on the real dataset of Meetup, the most popular event-based social network platform. Meanwhile, various metrics are explored to compare the recommendation methods. The experimental results show that the proposed algorithm outperforms the comparative ones.