AccessNet- Enhancing Public Transport Accessibility for People with Motor Disabilities

Public transportation networks result in a considerable amount of operational data; however, most cities do not have a smarter solution for the forecast and mobility planning of passengers. In this paper, we propose a novel AI- based framework called AccessNet for GTFS data-based public transportation demand prediction, which leverages historical ticketing data and temporal data. The framework preprocesses stop-level and route-level data, builds a graph model of the transit network, and uses a Long Short-Term Memory model for precise time-series-based forecasting. A comparative evaluation is also conducted using a baseline Exponential Moving Average model. Moreover, real OpenStreetMap data about roads is used with OSMNX, allowing for a realistic route geometric model, and a Folium-based interactive map is used to visualize the results of demand prediction, which represents congestion levels on the routes and results in a color-coded map of routes suitable for data-driven mobility planning and management of passengers on routes. Experiments show that our model outperforms other time-series-based models regarding temporal dependencies and demand changes over time.