In this paper, we describe a data-driven approach to the irregular flight recovery problem. By imitating the decision process of dispatchers, we develop a quantitative mechanism to evaluate the disrupted flight and to provide easily adopted recovery suggestions. Our method consists of two steps. The first step is to establish a scoring system based on interviews, questionnaires, and operational data. Specifically, an irregular flight will be assigned a real-time score representing the importance of the flight and the impact of its current status. This score is also used to evaluate the necessity of the immediate recovery of the irregular flight. The second step is to generate feasible adjustment plan, which decrease the total scores of the related flights. Having validated by 306 manually recorded recovery actions, the scoring system successfully explains most of the recovery actions from dispatchers, meaning that the scoring system is consistent with the airline recovery strategy. To further demonstrate the recovery method’s feasibility and response time, we also conducted tests based on one-day flight schedule simulation containing 92 flights and 7 historical cases from one Chinese airline. These tests prove the feasible adjustment plans can be generated in real-time, help airlines mitigate disruption effects to the network and reduce their decision process by 5–10 min in each delay scenario.