FUND-Trafikverket-RailForecast-2022

Utvärdering av ankomstprognoser för tåg (Evaluation of train arrival forecasts)

Today, there are several different ways to calculate train arrivals, such as manual calculations, automatic calculations based on digital graphs and AI solutions such as AIRT and TrainBrain. But how good are the calculations and in which situations is one or the other most reliable? This project proposal aims at comparing and evaluating different methods for train arrival estimates with the objective of increasing knowledge about the strengths and weaknesses of the different methods, for example in which situations the methods provide the most reliable estimates. We compile forecasts from a variety of sources, and evaluate them using both established methods and customised metrics developed in collaboration with traffic managers, informers and others in the industry. The aim is to help the Train Forecasting capability to select the best forecast model under the circumstances, thereby providing the best possible information to travellers and others, and ultimately delivering more satisfied travellers and reduced uncertainty during disruptive situations.