关键词: 空中交通管理, 飞行时间预测, 机器学习, 轻量梯度提升机（LightGBM）

Abstract: Accurately predicting terminal area flight time for inbound flights is of great significance for improving airport airspace operational efficiency and optimizing runway resource allocation. However, terminal area flight time is subject to the coupled effects of multi-source heterogeneous factors, including flight operations, air traffic flow, and weather. Traditional prediction models struggle to fully capture its nonlinear characteristics, and there is considerable room for improving prediction accuracy. To address this problem, this study first statistically analyzes the distribution characteristics of terminal area flight time for inbound flights and extracts 15 key influencing features from three dimensions: flight operations, air traffic flow, and weather. A terminal area flight time prediction model based on the light gradient boosting machine (LightGBM) is then constructed to learn the complex mapping between features and flight time, thereby achieving accurate prediction. Using actual operational data from Beijing Daxing International Airport as an example for simulation verification, the experimental results show that the proposed model achieves prediction accuracies of 68.25%, 91.32%, and 96.33% within ±1, ±3, and ±5- minute prediction windows, respectively, significantly outperforming traditional machine learning methods such as random forest and extreme gradient boosting (XGBoost). This study shows that the LightGBM model can effectively integrate multi-source heterogeneous features to achieve high-precision prediction of terminal area flight time for inbound flights, offering reliable data support for terminal area traffic management and runway operating mode optimization.

Key words:  air traffic management, flight time prediction, machine learning, light gradient boosting machine (LightGBM)