Research on Air Traffic Situation Prediction Methods in Multi-Airport Terminal Areas

Accurate assessment and forecasting of air traffic conditions in multi-airport terminal areas are essential for improving early-warning capabilities, mitigating flight conflicts, and alleviating air-route congestion. Accordingly, this study develops an air traffic situation prediction approach that combines an air route–flight state interdependent network with an Optimal Training Sample Online Fuzzy Least-Squares Support Vector Machine (OTSOF-LSSVM). An interdependent network model is first established. The Analytic Hierarchy Process (AHP) is then employed to combine three network indicators, namely node degree, weighted clustering coefficient, and node strength, thereby producing a comprehensive air traffic situation value and its corresponding evolutionary time series. Considering the time-varying and long-periodic properties of this series, an OTSOF-LSSVM-based prediction method is developed. Training samples are selected according to their temporal and spatial proximity to the prediction moment. In addition, block-matrix operations are introduced during model updating to streamline the computational procedure and improve algorithmic efficiency. The proposed approach is validated using actual flight data from the multi-airport terminal area of the Guangdong–Hong Kong–Macao Greater Bay Area. The results demonstrate that the proposed assessment method can effectively characterize the prevailing air traffic situation. Moreover, in comparison with several existing prediction techniques, the proposed method achieves the best overall performance, yielding a mean absolute error of only 0.0111.