USER ATTRIBUTE LEARNING FROM SOCIAL MEDIA AND UBIQUITOUS SENSORS

Abstract

The huge amount of user-generated data generated from users' online behaviors and ubiquitous sensors contains rich information about users' various knowledge and they are crucial for many useful applications like personalization, advertising and recommendation. In this context, we try to depict either the digital identity or the physical identity of a given user. This thesis first investigate the user physical attribute learning and focus on the case study of user activity attribute learning from multiple sensors. Then we study the user dynamic digital attribute learning and targets at the case study of user occupational attribute learning from multiple social networks, which can be easily generalized to other dynamic digital attribute learning scenarios. Finally, we also present a novel multi-task learning framework to learn user political attributes as a case study of static digital attribute learning. The experimental results on several real-world datasets have demonstrated the effectiveness of our approach.