Algorithm A1. Estimation procedure for the proposed NCGM.
Require:	Spatio-temporal population data $Y$ location information ${\left\{x_t\right\}}_{t=1}^L$ neighbor information ${\left\{N_l\right\}}_{\mathrm{l}=1}^{\mathrm{L}}$ hyperparameter $\lambda$
Ensure:	Population flow $Z$ estimated neural network parameters $\phi ;$ Repeat Calculate transition probability ${\theta }_{tl}$ by neural network $f$ for $t=1$ to $T-1$, $l=1$ to $L;$ Calculate the objective function $G$ and its gradient with respect to $Z$ and $\phi ;$ Update $Z$ and $\phi$ using gradient; until End condition is satisfied