A Data-Driven Heuristic Method for Irregular Flight Recovery

In this study, we develop a data-driven heuristic method to solve the irregular flight recovery problem. Based on operational data from China South Airlines, Beijing, China, we evaluate the importance of a flight in the flight network and the influence of a delay on a flight and its subsequent flights. Then, we classify historical states into three scenarios according to their delay reasons and investigate the recovery patterns for each scenario. Inspired by the results of the data analysis, we develop a heuristic algorithm that imitates dispatcher actions. The algorithm is based on two basic operations: swapping the tail numbers of two flights and resetting their flight departure times. The algorithm can provide multiple recovery plans in real time for different scenarios, and we continue to refine and validate the algorithm for more robust and general solutions through a cost analysis. Finally, we test the efficiency and effectiveness of the recovery method based on the flight schedule, with real and simulated delays, and compare it with two other methods and the recovery actions of dispatchers.