The Use of Graph Neural Networks in Rail Transport Planning

This work explores how graph theory concepts and neural networks can assist in strategic planning of metro network expansions using publicly available city data with the São Paulo Metropolitan Region. The methodology consolidated information from multiple public sources, developed a formula to estimate passenger demand based on catchment areas, applied Random Forest to identify the most relevant demographic features, and implemented a GraphSAGE model for demand prediction, extracting predictive capability from the topology of the system as well as socioeconomic features and Origin-Destination trips. The model achieved an R² of 0.874 ± 0.042 with minimal overfitting, outperforming the Random Forest approach. It can be applied to predict demand for future projects in a way that is accurate, inexpensive and computationally efficient, while also not requiring any rail system specific information besides topology. In this project, it was used to analyze multiple real projects and proposals for the São Paulo Metropolitan Region. Analysis revealed that employment, residences, and destinations where people go to eat represent 65% of the predictive capacity in the city.