Delays on German trains have become something of a running joke. Frequent travellers on the national Deutsche Bahn service often brace themselves for a delay announcement the moment they board. Deutsche Bahn runs the majority of the country’s railways, including 95 percent of long-distance routes, 67 percent of local transit, and 42 percent of freight transport. Yet in 2023, only 64 percent of its long-distance trains arrived on time—a punctuality benchmark defined as within six minutes of schedule. Worse, these figures exclude severely delayed or cancelled services.
According to a 2023 study by the nonprofit Consumer Choice Center, six of Europe’s 10 least passenger-friendly stations were in Germany. The index, which ranked stations on factors like connectivity and punctuality, highlighted Germany’s poor performance compared to its European neighbours. Meanwhile, the roads aren’t much better; in 2023, Germany endured a staggering 877,000 kilometres of traffic jams. That’s where Bahnvorhersage comes in. This service is designed to help users identify the most reliable train connections.
<!–
window.CustomSubstackWidget = {
substackUrl: “https://digitalforgood.substack.com/”,
placeholder: “my email address”,
buttonText: “Subscribe”,
theme: “custom”,
colors: { primary: “#57B050”, input: “#000000”, email: “#FFFFFF”, text: “#000000”,
}
};
–>
Developed by two students from Tübingen, this innovative project utilises an AI model trained on approximately 700 GB of railway connection data. The system predicts the likelihood of passengers making one or more connections, offering valuable insights for journey planning. According to Theo Döllmann, who presented the project at the 38C3 Hacker Congress, the predictions closely align with Deutsche Bahn’s actual delay patterns, making it a reliable tool for navigating Germany’s often unpredictable train services.
How Bahnvohersage works
Bahnvorhersage is a free tool designed to mirror the interface of the DB Navigator. It allows travellers to input their departure station, destination and travel time as usual, however, users can also filter their search to focus on local services or trains that allow bicycles.
While the tool displays the same connections as DB Navigator, it offers something extra: probabilities for making each required transfer. This added insight helps travellers plan their journeys with less guesswork, and therefore stress.
We gave it a go. For a journey from Berlin Central Station to Eisenhüttenstadt, a town on the Polish border around 100 kilometres away, there was a connection available at our chosen time with one transfer. This was to take just under two hours. The system calculated a 95 percent probability of successfully making all connections on this route. However, the next available connection had a much lower success rate, at just 50 percent.
A key issue lay with a regional train which often arrives in the connecting station a few minutes late. With only a five-minute transfer window, the likelihood of missing the connection is high, making delays almost inevitable on this route.
The AI developed for Bahnvohersage takes several factors into account for its calculations. According to Theo Döllmann, in addition to the time and train type, the route already travelled and the time of the forecast are also important for the analysis. The model uses these factors to calculate a transfer score that matches the actual delay data collected by Deutsche Bahn with precision.
But back to our example: there was a 93 percent probability that we would arrive on time via the first connection. Experienced rail travellers would have chosen the route with fewer changes anyway. However, the system is a good help, especially for longer train journeys along unfamiliar routes. Long-distance trains to other countries in particular are much easier to book with digital tools.
Open data set enables further analyses
However, Bahnvohersage has added value due to the way in which collected data is handled. They make the one billion connections they have collected for analysis freely accessible in accordance with the open data principle.
Public rail transport is an important part of the mobility transition all across Europe and the world. Germany is no different. In recent years, there’s been more and more pressure on the car-loving nation to reduce emissions from traffic and car use. The fact that Völlmann and De Kuthy Meurers are making the data freely available allows political parties, organisations and associations to formulate and illustrate arguments for the renovation and expansion of the rail network.
Railway forecast seeks feedback for new function
Bahnvorhersage.de is currently offering a new function as an alpha test. Interested parties can use the service’s top navigation to find alternatives to their connections in advance. The platform wants to offer these alternatives at the same time as the search. To do this, it wants to deviate from the railway’s connection search and replace it with its own system.
Although the service is not yet reliable, the feedback helps the developers to find bugs. Anyone who wants to can still test the service and support the project. Feedback is provided via a project page at the provider GitLab.
The post Missed Connection: Bahnvohersage Uses AI & Open Data to Predict Train Delays appeared first on Digital for Good | RESET.ORG.