GLORY: Exploration and integration of global and local correlations to improve personalized online social recommendations

Abstract

Nowadays people manage their social circles via a variety of online social media which employ social recommendation as an important component. Among social recommendation methods, global methods take an emphasis on common tastes between people while local methods assume that new relations are established mainly through people's common friends. However, in a real social network, both local and global relations exist, which motivate us to integrate them to improve recommendation performance. To achieve the goal, we proposed a novel hybrid method GLORY to combine global associations with local correlations for social recommendation. GLORY consists of two components: GLOBE and LORY. The former is a globalised regression model to explore the concordance between people's preference with the relatedness of their friends. The latter is an integration method to fuse global and local correlations via a rigorous statistical model to calibrate the statistical significance of these correlations. Furthermore, we demonstrated the effectiveness of our methods via 10-fold large-scale cross-validation on three real social network datasets (Facebook, Last.fm and Epinions). Results show that GLORY significantly outperform the state-of-the-art methods while LORY is effective across various global and local methods, indicating their promising future for social recommendations.

Blazing the Trail: Considering Browsing Path Dependence in Online Service Response Strategy

“…For the field of personalized recommendation, current research usually studies how to improve the accuracy of personalized recommendation from the perspective of algorithms (Chang & Jung, 2017;Gan et al, 2019). One of the biggest challenges for recommendation algorithms is data sparsity and the cold start problem (Bunnell et al, 2020).…”