Time-Dependent Origin-destination Estimation from Land Use and Mobile Phone Data

Mobile network data have been increasingly applied to measure the spatiotemporal changes in population due to its cost-benefit compared with conventional survey methods. Furthermore, as one of the essential aspects of urban areas, land use plays a vital role in transportation planning. With these two data comes the opportunity to improve understanding human mobility patterns. This work utilizes mobile-phone-based OD-matrices and land use data to predict the travel demand for the province of Liège in Belgium. Hourly mean OD-flows are available at the sub-commune level in the research area. In the process, we first identify the relationship between land use and the movements of individuals derived from mobile phone data over a typical week and a holiday week. It is shown that the observed number of trips and land use are closely linked at the aggregation level. After that, we deploy neural network models to estimate OD-flows considering land use features, the spatial correlation of traffic analysis zones, and the temporal correlation of time series in existing flows. The study aims to explore the effect of land use on human mobility patterns captured by mobile phone data while offering a feasible solution to predict time-dependent OD-flows based on these two data.