Find you from your friends: Graph-based residence location prediction for users in social media

As a bridge between social media and physical space, location information will potentially make the internet smarter, and release the real power of social media to address the serious and significant problems in the real world. However, in terms of privacy and security, most of the users are unwilling to make their locations public. To address the problem, an algorithm is necessary to predict the users´ residence locations based on the public profiles. We define location propagation probability of users, leverage a semi-supervised learning algorithm, and introduce a novel method of location propagation to predict users´ residence locations based on users´ social relationships, textual and visual contents and a small amount of known users´ residence locations. The experimental results on a large scale real data set in Tencent Weibo demonstrate that our location propagation algorithm outperforms the state-of-the-art approaches in both accuracy and scalability.