Scholarlink

With the rapid development of the global aviation industry, reducing flight delays has become a critical challenge. Aircraft taxi-out time, which directly impacts the utilization of runways and taxiways, is one of the most significant factors influencing the operational efficiency of airport field operations. This study takes Beijing Capital International Airport as a case study, identifying model characteristic parameters through a comprehensive analysis of historical departure data. An integrated long short-term memory–extreme gradient boosting (LSTM-XGBoost) model is proposed to predict taxi-out times, with further optimization achieved using the sparrow search algorithm (SSA). To evaluate the model’s generalizability, historical data from Zhengzhou Xinzheng International Airport are utilized for validation. Comparative analysis of prediction evaluation metrics reveals that the integrated LSTM-XGBoost model outperforms standalone LSTM and XGBoost models, demonstrating superior prediction accuracy and robustness.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of Taxi-Out Time Based on LSTM-XGBoost

  • 章节

  • 发表日期：

    2026年

查看全文
  • Guoxin Zhang,
  • Guan Lian,
  • Benxiao Lou,
  • Wenyong Li

摘要

With the rapid development of the global aviation industry, reducing flight delays has become a critical challenge. Aircraft taxi-out time, which directly impacts the utilization of runways and taxiways, is one of the most significant factors influencing the operational efficiency of airport field operations. This study takes Beijing Capital International Airport as a case study, identifying model characteristic parameters through a comprehensive analysis of historical departure data. An integrated long short-term memory–extreme gradient boosting (LSTM-XGBoost) model is proposed to predict taxi-out times, with further optimization achieved using the sparrow search algorithm (SSA). To evaluate the model’s generalizability, historical data from Zhengzhou Xinzheng International Airport are utilized for validation. Comparative analysis of prediction evaluation metrics reveals that the integrated LSTM-XGBoost model outperforms standalone LSTM and XGBoost models, demonstrating superior prediction accuracy and robustness.