Mobility Knowledge Discovery to Generate Activity Pattern Trajectories

Geographic mining of mobility behavior of individuals and analysis of multi-dimensional correlations between travelers, locations, and activities is crucial in the design and operation of urban infrastructures. The work presented here develops methods to learn mobility behavior of travelers through a mobility knowledge discovery process and to infer the spatial-temporal attributes of activity trajectories based on the demographics of travelers and the features of urban forms. The proposed mobility knowledge discovery process is comprised of two sections: (a) modeling the transactions between activity state vectors (type, location, duration and time), and (b) modeling the correlations between travelers and locations. After activity trajectories are generated, the inferred patterns are assigned to census tracts using an OD-based probabilistic forward-backward method. The proposed methods are validated by applying the model on household travel survey data collected for New York City.