A young woman gets on a bus smiling at the driver, ready to validate.
Projects

Urban mobility: using AI to predict ridership based on weather conditions

Keolis Nederland
Share on
PostedAPR. 5, 2022
Words byKeolis
Partners
    Weather can have a big impact on public transit demand. For this reason, Keolis Nederland has adopted an urban mobility solution that uses artificial intelligence (AI) to predict ridership levels based on weather conditions. This information makes it possible to improve resource allocation and design more resilient networks.

    Context: determining ridership as the weather changes

    Weather often has a big impact on bus ridership. In the Netherlands, bad weather often leads to a significant increase in public transport demand, especially on workdays. For this reason, Keolis Nederland sought to develop an urban mobility solution for accurate predictions of passenger flows, in order to improve service and optimize the bus network. 

    Innovation: using artificial intelligence to predict ridership

    The solution? Artificial intelligence. Keolis Nederland used a deep learning algorithm to analyze its bus network in the region of Twente. By crossing smart card data from riders with weather conditions surrounding the bus departure times, the AI system provides the public transport provider with more accurate predictions of ridership on its bus lines. With this information adapted to current and future weather conditions, Keolis Nederland can better allocate bus resources and design more resilient bus networks.

    Photo montage of a bus with the inscription "AI in Passenger Prediction
    Ai in Passenger Prediction

    +66,4%

    Thumb ok

    Improved accuracy of predictions for weekend days

    +27,4%

    Thumb ok

    Improved accuracy of predictions for week days

    Value: optimizing transit networks with accurate ridership predictions

    Throughout its initial round of testing, the solution showed high promise in its ability to accurately predict bus ridership based on weather conditions. Integrating weather data into the new AI system improved its prediction accuracy by 27% on weekdays and by 66% on weekends compared to its accuracy without weather data. By combining historical passenger data with meteorological information, Keolis Nederland can obtain a realistic forecast of bus ridership. With more accurate insights into future passenger flows, the operator is able to optimize its bus network in the city. 

    Next steps: expanding the solution to cover new lines and fields 

    Following the promising results of this initial experiment, Keolis can now leverage its data volumes to adapt and expand this urban mobility solution throughout the rest of the Group. The new AI system is highly replicable and can be easily adapted to other networks. Moreover, the deep learning principle behind this solution can be adapted to many other applications, such as passenger information, revenue prediction, tendering and more. 

    Image of neural networks made by computer.
    Neural networks
    A red bus from the region of Twente is driving on a road.

    © Keolis Nederland

    Keep up to date with all the latest innovation news from the Keolis group by subscribing to our newsletter.