evaluate if is a good idea to use the Deep Gravity model for mobility flows generation

[napo]

A Deep Gravity model for mobility flows generation
https://www.nature.com/articles/s41467-021-26752-4

[napo]

I also suggest reading the paper on the Radiation model by Filippo Simini

Usually on city scale the radiation is better but it has the limit of being parameter free while the gravity is more manageable. In both cases, it would be good to have flows at least for training (the whole city is not needed, just a part). In general, both use only population and distance as features

Both models are implemented in scikit-mobility