<i>TiI</i> : A Novel Framework for Predicting Pre-Tactical Multiple Airports Traffic Flows by Integrating Flight Plans and Weather Data

The accurate prediction of traffic flows, defined as the aggregate forecasting of combined departure and arrival volumes across multiple airports within a designated air traffic control zone during the pre-tactical phase, is crucial for effective air traffic flow management, particularly given the rapid growth of aviation transportation. However, existing methods face significant challenges in modeling the complex spatial-temporal dependencies due to difficulties in constructing appropriate graph structures, as well as limitations inherent in the algorithms and periodic trend-based prediction mechanisms. To address these challenges, a novel framework termed Transformation into Image (TiI) is proposed in this study. Initially, data from multiple airports are transformed into images, effectively compressing spatial-temporal features into grid-based distribution representations. Subsequently, the proposed multi-kernel convolution module and parallel patch-transformer module are integrated into each resolution stream of the U-Net architecture to accurately capture spatial-temporal dependencies at both coarse-grained and fine-grained levels. Finally, the predicted images are converted and partitioned into traffic flow data for individual airports, yielding the desired results. To the best of our knowledge, TiI represents the first approach to synchronously extract both spatial and temporal dynamics by encoding traffic flow data into image representation. Extensive experiments validate the feasibility of TiI, demonstrating that it significantly outperforms comparative methods. Furthermore, an ablation study confirms the effectiveness of the two core components within the TiI regression model.