Data-Driven Departure Flight Time Prediction Based on Feature Construction and Ensemble Learning

Temporal–spatial resource optimization within the terminal maneuvering area has become an important research direction to meet growing demand for air traffic. Accurate departure flight time prediction from taking off metering fixes is critical management, connecting surface operations, and overhead stream insertion. This paper employs ensemble learning methods (including bagging, boosting, stacking) predict times via different based on four feature categories: initial states, operating situation, traffic demand, wind velocity. The stacking method a linear regressor, support vector tree-based regressor as base learners. Guangzhou Baiyun International Airport case study shows that proposed in this work outperforms other could achieve satisfactory performance prediction, with high accuracy of up 89% 1 min absolute error 98% 2 error. Besides, affecting factors analysis indicates operation direction, distance, areas significantly improve accuracy.