Urban region representation learning with human trajectories: a multi-view approach incorporating transition, spatial, and temporal perspectives

Today's article comes from the journal of GIScience & Remote Sensing. The authors are Zhang et al., from Shandong University, in China. In this paper they use anonymized data from cell-towers to create geo-spatial and temporal views of a major city. Then they used those views to create a Graph Neural Network that could infer population density, housing prices, and land-use classification.

DOI: 10.1080/15481603.2024.2387392