Urban Knowledge Graph Aided Mobile User Profiling

Nowadays, the explosive growth of personalized web applications and rapid development artificial intelligence technology have flourished recent research on mobile user profiling, i.e., inferring profile from behavioral data. Particularly, existing studies mainly follow data-driven paradigm to develop feature engineering representation learning such data, which however suffer robustness issue, generalizing poorly across datasets profiles without considering semantic knowledge therein. In comparison, rising knowledge-driven built upon graph (KG) offers a potential solution mitigate weakness. Therefore, in this paper, we propose Knowledge Graph aided framework for Mobile User Profiling (KG-MUP). Specifically, distil among firstly construct an urban (UrbanKG) with domain entities like users, regions, point interests (POIs), etc. identified, as well relations home, workplace, spatiality, extracted. Moreover, leverage tensor decomposition neural network obtain knowledgeable representations UrbanKG. addition, introduce several customized features quantify individual mobility characteristics profiling. Extensive experiments three real-world demonstrate that KG-MUP achieves state-of-the-art performance inference tasks. further results also reveal importance various inference, provides meaningful insights modeling