Abstract
The movements of individuals within and among cities influence critical aspects of our society, such as well-being, the spreading of epidemics, and the quality of the environment. When mobility data are not available for a particular region of interest, researchers must rely on mathematical models to generate mobility flows. Here we propose the Deep Gravity model as an effective method of flow generation that exploits many variables (e.g., land use, road network, transport, food, health facilities) extracted from voluntary geographic data and uses deep neural networks to discover non-linear relationships between those variables and mobility flows. Our experiments, conducted on commuting flows in England, show that Deep Gravity has good geographic generalization capability, achieving a significant increase in performance (up to 350% in densely populated regions of interest) with respect to the classic gravity model and models that do not use deep neural networks or geographic data.