Modeling sector route segment capacity under convective weather based on machine learning

Abstract Estimating airspace capacity under convective weather conditions is crucial for ensuring the efficiency and safety of air traffic operations. Sector route segments, which are essential components of flight routes, require timely capacity predictions during operationally critical periods. In this paper, initially, an enhanced Recursive Feature Elimination algorithm is used to select meteorological data and develop predictive features. Subsequently, the CWSRC model is established using the RF supervised learning algorithm. Finally, the paper takes ENH-YIH segment as an example to predict the capacity. Compared with other machine learning algorithms, the residual percentages for KNN, MLP and RF are 86.03%, 77.37% and 93.40%, respectively, within the range of [−0.2, 0.2]. In three separate day cases, results show that the CWSRC model’s MAE, MSE, RMSE and R 2 significantly outperform traditional methods like Maxflow/Mincut and scanning line. The results confirm the CWSRC model’s superior predictive capabilities.