From Prediction to Assignment: A Statistical Learning Framework for Robust Airport Gate Assignment Under Uncertainty

ABSTRACT

Airport gate assignment is highly sensitive to flight arrival uncertainty, as deviations between scheduled and actual arrival times can easily disrupt pre‐planned gate assignments and reduce operational efficiency. To address the gap between arrival time prediction and gate assignment, this paper proposes an integrated prediction‐assignment framework. In the prediction stage, an ITLR model combining Transformer, LSTM and Random Forest (RF) with improved grey wolf optimization (IGWO) is developed to predict flight arrival times from cleaned operational data. In the assignment stage, the predicted arrivals are embedded into an adaptive diversity‐aware multi‐objective genetic algorithm (ADA–MOGA) to balance bridge utilization and fuel consumption. A case study using operational data from Zhengzhou Xinzheng International Airport shows that ITLR achieves an R ² of 0.9495. In the detailed assignment case, the proposed framework reaches 92.5% bridge utilization and reduces fuel consumption relative to benchmark methods. Further tests on flight instances of different scales show that the proposed framework maintains stable advantages over optimization‐based and learning‐based baselines. The results indicate that flight‐specific arrival prediction can improve gate‐assignment efficiency and operational flexibility under uncertainty.