基于XGBoost-MLP集成方法的离港航班延误预测  

An integrated prediction method for departure flight delay based on XGBoost and MLP

作　　者：张铭梁 侯霞[1] ZHANG Mingliang;HOU Xia(Computer School,Beijing Information Science&Technology University,Beijing 100101,China)

机构地区：[1]北京信息科技大学计算机学院,北京100101

出　　处：《北京信息科技大学学报（自然科学版）》2022年第3期41-45,共5页Journal of Beijing Information Science and Technology University

基　　金：国家自然科学基金资助项目(61672105)。

摘　　要：为了更准确地描述航班延误情况,为旅客出行提供参考,使用极端梯度提升(extreme gradient boosting, XGBoost)算法与多层感知机(multilayer perceptron, MLP)集成的模型对离港航班延误状态进行预测,将传统的延误、不延误细分为延误、半延误和不延误3种情况。在对航班数据和天气数据进行合并、筛选、拆分的基础上,先基于XGBoost模型进行二分类预测,然后基于二分类结果使用MLP进行三分类预测。实验结果表明,该方法比仅使用XGBoost模型或者MLP模型预测效果更佳,并且可改善半延误区间误差高的问题。In order to describe the flight delay situation more accurately and provide reference for passengers′ travel, a prediction method integrating XGBoost model and MLP model was used to predict the departure flight delay status, and subdivide the traditional delay and no delay situations into three situations: delay, semi-delay and no delay.On the basis of merging, screening and splitting flight data and related weather data, the binary-classification prediction was made based on XGBoost model, and then the tripartite-classificatim prediction was made based on the predicted dichotomous results using MLP.Experimental results show that this method is better than XGBoost model and MLP model, and can help to solve the problem of high error in semi-delay part.

关 键 词：极端梯度提升(XGBoost) 多层感知机(MLP) 多分类 集成方法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]