A Multi-Airport Flight Schedule Reallocation Algorithm Based on Unbiased Meta-Learning

To address unforeseen disruptions, the flight schedules reallocation at the pre-tactical or tactical stage can effectively prevent potential delays and ensure flight safety.The traditional allocation algorithms are generally cost-minimization-oriented and fail to account for the specific impacts of disruptions on individual flights.This paper tackles the challenge posed by limited samples of disruption scenarios by constructing a comprehensive set of allocation factor indices that account for multiple influencing factors.To mitigate the effects of regional differences in disruption probabilities on algorithm generalization, compressed sensing technology is employed for feature selection.Furthermore, an unbiased meta-learning-based flight schedule reallocation algorithm is proposed.The experimental results indicate that the entropy-integrated algorithm achieves a maximum classification accuracy of 97.01% at a learning rate of 0.03.The predictive accuracy of the trained multi-airport flight schedule reallocation models for allocation levels exceeds 95%.