Stockholmståg, the train operator in the Swedish capital of Stockholm, is testing a system that can accurately predict train delays before they happen.

The company has built a prediction model, using

big data, which provides a view of the entire commuter train system two hours into the future.

“We can now forecast disruptions in our service and our traffic control centre can prevent the ripple effects that actually cause most delays,” said Stockholmståg communications director Mikael Lindskog.

Currently still in test phase, the forecasting model, known as the “commuter prognosis”, is based on a mathematical algorithm developed by Stockholmståg and data scientist Wilhelm Landerholm.

Key to the model is a sufficient amount of historical data. The model works similarly to a seismograph, but instead of earthquakes it identifies late train arrivals. When this happens, the algorithm uses historical data from previous occurrences to forecast the impact this will have on the entire train network.

Real-time public transportation information is already used around the world, but according to Lindskog, traffic control centres typically analyse any delays manually to try to prevent further effects. But the commuter prognosis system will forecast these delay effects automatically and show how a current disturbance will affect the train network in the near future.

Read more about IT in Sweden

For example, if the system predicts a train will be 10 minutes late arriving at a certain station in two hours, the traffic control staff should be able to minimise the delay and its impact by asking the train to skip a stop or by issuing extra trains. The algorithm will then detect the effect this has and re-calculate its forecast for the entire train network.

“The time it takes to run a new forecast in the system for the next two hours is less than one minute,” said Landerholm. “If you change something [such as deploying a new train], you will see its effects quickly in the system.”

App in the pipeline

While Stockholmståg doesn’t yet have a timeline for implementing the commuter prognoisis model in its traffic control, Stockholm commuters can benefit from it later in 2015 when the train operator releases a smartphone app based on the model. The train travel app will also integrate other transportation data aimed to make commuting easier.

“The app will indicate which [coaches] are more or less crowded,” explained Lindskog. “We measure the wheel pressure in the trains [to estimate this].”

But this is just the start. According to Landerholm, the model can be used in any city as long as there is enough train network data. The approach could also be expanded outside train networks to wherever there are regular routes and schedules, such as busses or even delivery routes.

“The commuter prognosis will be the first automated forecasting model of its kind. In a long-time perspective it’s possible it will change how traffic control centres all over the world work,” Lindskog said.