T-Systems
Artificial intelligence for Deutsche Bahn
Deutsche Bahn is planning to improve the forecasting of train arrival and departure times. In future, data analytics services from T-Systems will be used in its long-distance, regional and suburban rail services, which will continuously learn from real-time forecasts.
To do this, T-Systems compares the timetable data for more than two million stops per day for all timetabled passenger transportation with the current traffic situation every minute. This results in a real-time forecast of the expected arrival time and its impact on possible connections. During the calculations, the position reports of all moving trains are analyzed within seconds in T-Systems' data centers and a forecast for the rest of the train run is created in real time. The algorithm used for this is 'intelligent' - in other words, it uses machine learning methods, among other things. It uses various forecasting models, which are selected dynamically depending on the traffic situation. The models are trained at night on a 24-hour basis using historical data. This continuously improves the accuracy of the forecast and adapts it to the current traffic conditions. The forecasting solution is based on an in-house development by T-Systems and its subsidiary T-Systems Multimedia Solutions and is being further developed and introduced in a joint project with Deutsche Bahn.
From the second quarter of 2017, the solution will noticeably improve the information provided to passengers in the event of delays to long-distance and local trains and will then be continuously enhanced with additional data - in line with the concept of smart data. Passengers will be able to find out about departure times in real time up to 90 minutes in advance via smartphone and app, as well as at train stations.










