基于GBDT的多时刻航班过站关键节点时间预测模型  

Multi-Time prediction model of flight transit key nodes time based on GBDT

作　　者：丁建立[1] 冯昊 DING Jianli;FENG Hao(College of Computer Science and Technology,CAUC,Tianjin 300300,China)

机构地区：[1]中国民航大学计算机科学与技术学院,天津300300

出　　处：《中国民航大学学报》2024年第6期27-33,共7页Journal of Civil Aviation University of China

基　　金：国家自然科学基金项目(U2033205)。

摘　　要：为准确预测离港、起飞等航班过站关键节点时间,提高繁忙机场运行效率,本文提出一种基于梯度提升决策树(GBDT,gradient boosting decision tree)的多时刻航班过站关键节点时间预测模型。首先,按产生时刻划分航班信息数据项类别;其次,基于GBDT算法和Spark平台分别构建不同过站时刻的航班过站关键节点时间预测模型;最后,以实时计算方式获取和处理航班数据,实现在多个时刻对航班离港时间和起飞时间进行动态预测。实验结果表明,该模型具有良好的预测表现,并与其他算法进行对比,预测效果最优,±15 min内预测准确率达到95.6%。To accurately predict the flight transit key nodes time such as departure and takeoff,and improve the operational efficiency of busy airports,a multi-time prediction model of flight transit key nodes time based on gradient boosting decision tree(GBDT)is proposed in this paper.Firstly,the flight information data items are classified according to the generation time.Secondly,based on the GBDT algorithm and Spark platform,the prediction models of flight transit key nodes time at different transit times are constructed respectively.Finally,flight data is obtained and pro-cessed in real-time computing manner,enabling dynamic prediction of flight departure time and take-off time at multiple times.The experimental results show that the proposed model has good predictive performance and has the best predictive performance compared to other algorithms,with a prediction accuracy of 95.6%within±15 minutes.

关 键 词：航班过站关键节点时间 梯度提升决策树(GBDT) Spark平台 动态预测 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]