ATD: Statistical Modeling of Spatial Temporal Human Mobility Flows from Aggregated Mobile Phone Data

The widespread usage of smartphones and GPS-enabled applications has led to an increase in the availability of movement data products. An example of such data is the spatial temporal human mobility flow, which captures the movement patterns of individuals from specific origin locations to destination locations over time. Learning the patterns of human mobility flow in a normal context could provide valuable information for applications such as urban and traffic planning. Furthermore, analyzing the impact of events such as crises on mobility patterns can aid in the deployment of early interventions and responses. In this project, the investigator will study community detection problems of spatial temporal human movement origin-to-destination (OD) flow networks, with a focus on asymmetric flows and evolving community structures. The project will also provide training to graduate students involved in the research. In this project, the investigator will develop a new class of Bayesian random graph partition prior models that take into account spatial structures and contiguity constraints for clustering while allowing the number of spatial clusters to be unknown. The investigator will use this method to build a flexible Bayesian hierarchical stochastic block model for the spatial OD flow networks to detect origin and destination communities. Furthermore, the investigator will design an efficient Bayesian algorithm for the estimation of community structures together with uncertainty measures. Finally, the investigator will extend the model to a dynamic setting for the detection of temporally varying community structures while accounting for temporal dependence between community structures. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.