基于集成学习的离港航班延误预测方法  

Prediction of departure flight delay based on ensemble learning

作　　者：罗杰 侯霞[1] 杨鸿波[2] 刘林[3] 谢丰[3] LUO Jie;HOU Xia;YANG Hong-bo;LIU Lin;XIE Feng(School of Computer,Beijing Information Science and Technology University,Beijing 100101,China;School of Automation,Beijing Information Science and Technology University,Beijing 100192,China;System Evaluation Office,China Information Technology Security Evaluation Center,Beijing 100085,China)

机构地区：[1]北京信息科技大学计算机学院,北京100101 [2]北京信息科技大学自动化学院,北京100192 [3]中国信息安全测评中心系统评估处,北京100085

出　　处：《计算机工程与设计》2022年第4期1145-1151,共7页Computer Engineering and Design

摘　　要：为解决稀疏数据对预测模型带来的负面影响,提高以机场为主体的离港航班延误预测效果,提出一种基于Xgboost模型与Logistic模型相集成的离港航班延误预测方法。将Xgboost模型作为特征转换器,把森林中每棵决策树的叶节点作为新特征向量输入到Logistic模型中进行航班延误预测。通过在未经规范化的稀疏数据上和其它预测方法相比,该方法可以显著提高单独预测模型在稀疏数据集上的预测效果,相较于其它机器学习方法预测效果更佳。To eliminate the negative impact of sparse data on the prediction model and improve the airport-based departure flight delay prediction,a departure flight delay prediction method based on the integration of Xgboost model and Logistic model was proposed.The Xgboost model was used as a feature converter to input the leaf nodes of each decision tree in the forest as a new feature vector into the Logistic model for flight delay prediction.By comparing the method on un-normalized sparse data with other prediction methods,the method can significantly improve the prediction performance of the individual prediction models on sparse data sets,which presents better effects than other machine learning methods.

关 键 词：稀疏数据 特征转换 Xgboost模型 LOGISTIC模型 集成学习 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]