neuroGravity: AI Reconstructs Human Mobility Networks from Limited Data

A team of researchers has introduced neuroGravity, an advanced deep learning model grounded in physics that reconstructs human mobility networks based on limited data, enabling its application to cities that have not been observed. By utilizing only the distribution of urban facilities and populations, the model's regional representations exhibit a strong correlation with socioeconomic factors and livability, serving as scalable alternatives to expensive travel surveys. The findings indicate that spatial income segregation plays a crucial role in the model's transferability, as mobility networks are reconstructed most effectively when the target cities have comparable segregation levels to the source. An index was created to measure this segregation, enhancing the prediction of transferability. This research addresses significant challenges in urban planning and public health, especially in underdeveloped areas lacking thorough travel surveys.