Quantile-based Machine Learning Predictions of Luggage Arrival Time at Airport Reclaim Areas: A Data-Driven Approach Using Operational Data from Munich Airport

Luggage waiting time in reclaim areas significantly influences the arrival passengers’ experience at airports. This study leverages data from Munich Airport to develop quantile- based machine learning models that predict the time from the arrival of the aircraft to the time the first and last bags appear on the baggage reclaim belt, based on timestamps manually recorded by ground handlers. The ultimate aim is to enhance the passenger journey by sharing these predictions in real time, enabling passengers to make decisions on the use of their time while waiting. To account for the fact that the timestamp of the first bag does not always reflect the actual start of the luggage delivery process, often marking the deposit of only a few bags before the main set, quantile predictions above the 50th percentile were used to intentionally overestimate luggage delivery times. This approach would also ensure that passengers experience luggage delivery earlier than the predicted time communicated to them, potentially enhancing their overall perception of service quality. A set of developed models shows encouraging performance, indicating potential for providing passengers with timely and realistic predictions of luggage delivery time through airport displays or mobile apps. The next phase of the project will involve validation activities and post-deployment surveys to evaluate potential improvements in passenger satisfaction. The work has been conducted as part of the PRELUDE project, one of the recent EUROCONTROL initiatives.