Social User Profiling: A Social-Aware Topic Modeling Perspective

Abstract

Social user profiling is an analytical process that delivers an in-depth blueprint of users’ personal characteristics in social networks, which can enable a wide range of applications, such as personalized recommendation and targeted marketing. While social user profiling has attracted a lot of attention in the past few years, it is still very challenging to collaboratively model both user-centric information and social network structure. To this end, in this paper we develop an analytic framework for solving the social user profiling problem. Specifically, we first propose a novel social-aware semi-supervised topic model, i.e., User Profiling based Topic Model (UPTM), which can reconcile the observed user characteristics and social network structure for discovering the latent reasons behind social connections and further extracting users’ potential profiles. In addition, to improve the profiling performance, we further develop a label propagation strategy for refining the profiling results of UPTM. Finally, we conduct extensive evaluations with a variety of real-world data, where experimental results demonstrate the effectiveness of our proposed modeling method.