Publication
Study summary
The study develops open machine-learning models for estimating commercial-flight actual takeoff weight, an important input for trajectory prediction and environmental assessment that is rarely available in open operational datasets. The analysis uses large-scale aviation data released for the EUROCONTROL Performance Review Commission Data Challenge, combining flight, trajectory and meteorological information for 369,013 flights in European airspace during 2022. CatBoost, LightGBM, XGBoost, artificial neural networks and an ensemble are evaluated for predictive accuracy, computational efficiency and practical applicability. The best-performing approach achieves a mean percentage error of 1.73%. The work also emphasizes reproducibility and open implementation, providing a transparent benchmark for aircraft-mass estimation from operational data.
Keywords: aircraft mass; predictive modeling; machine learning; air traffic management; open aviation data.
Recognition: The work is based on the solution developed by Team Tiny Rainbow from the Aeronautics Institute of Technology, which finished 2nd in the PRC Data Challenge 2024. Official rankings.
BibTeX
Please cite the published journal article through its DOI.