Abstract:The wide-spread use of positioning devices (e.g., GPS) has given rise to a mass of spatio-temporal trajectories, which enable us to mine user-mobility patterns. In this paper, we proposed a model based on sequential and temporal trajectory features to mine peoples latent movement patterns. Considering the following trajectory characteristics---(1) location sequences play a pivotal role in understanding user-mobility patterns, and (2) user-mobility patterns change over time---we first modeled the location sequences and then incorporated the temporal information into the model. To verify the effectiveness of our model, we performed thorough empirical studies on a check-in dataset of the Gowalla social network. The experimental results confirmed that the proposed method performed better than Latent Dirichlet Allocation (LDA) and T-BiLDA.

Mining mobility patterns based on temporal and sequential features