Discovering daily routines from Google Latitude with topic models

Discovering users' whereabouts patterns is important for many emerging ubiquitous computing applications. Life-log systems, advertisement and smart environments are only some of the applications that can be supported by information regarding user patterns and routine behaviors. Latent Dirichlet Allocation (LDA) is a powerful mechanism to extract recurrent behaviors and high-level patterns (called topics) from mobility data in an unsupervised manner. In this paper we test the effectiveness of LDA in identifying users' routine behaviors from mobility data collected with Google Latitude. Results show that the proposed technique provides good results in discovering patterns and routine behaviors.