Keywords
Prediction, RNN, Public Transport, PWA, Bus

Technologies
Django, Keras, Docker, PostgreSQL, PWA

"Ghost Buses" - scheduled buses that appear on the digital displays at bus stops but never actually arrive - are a frequent frustration for commuters in Ireland. GhostBUSter is a machine learning powered solution designed to eliminate this uncertainty by providing realistic predictions of when the bus will arrive.

The primary objective of the project was to develop an application that leverages National Transport Authority data to forecast bus arrivals around Cork. At its core, the system uses a neural network trained on historical real-time data to generate predictions. By filtering out buses that have been cancelled, deviated from their official route or lost tracking mid-transit, the application ensures that predictions are only made for buses likely to reach the user. The data is served via a Progressive Web App, offering the seamless experience of both a website and a dedicated mobile application. Ultimately, GhostBUSter gives regular commuters a simple, fast, and reliable way to check their bus arrival times without the usual uncertainty or hassle.